The mAP is always 0 when I train on my custom data in coco format? #504

Hezhexi2002 · 2021-08-15T14:30:38Z

I tried to train on my custom data in coco format but no matter how I train for 10 epochs or 300 epochs I still get mAP=0,my custom data is originally in YOLOV5 format so I use this https://github.com/RapidAI/YOLO2COCO to convert my YOLOV5 format data into coco format and I also check the label which is correct.Besides I also have changed the classes in yolox/data/datasets/coco_classes.py to my own classes.However the result disappointed me again,then I tried to use the datasets https://drive.google.com/file/d/16N3u36ycNd70m23IM7vMuRQXejAJY9Fs/view?usp=sharing you guys suggest in the train_custom_data.md,but for well-known reasons in China I can hardly download it,so I also suggest you provide a BaiduYun version which is more friendly to our students.
Here is my train_log.txt:
2021-08-15 21:54:11.090 | INFO | yolox.core.trainer:before_train:126 - args: Namespace(experiment_name='606', name=None, dist_backend='nccl', dist_url=None, batch_size=8, devices=0, exp_file='exps/example/custom/606.py', resume=False, ckpt='yolox_s.pth', start_epoch=None, num_machines=1, machine_rank=0, fp16=True, occupy=False, opts=[])
2021-08-15 21:54:11.091 | INFO | yolox.core.trainer:before_train:127 - exp value:
╒══════════════════╤════════════════════════════╕
│ keys │ values │
╞══════════════════╪════════════════════════════╡
│ seed │ None │
├──────────────────┼────────────────────────────┤
│ output_dir │ './YOLOX_outputs' │
├──────────────────┼────────────────────────────┤
│ print_interval │ 10 │
├──────────────────┼────────────────────────────┤
│ eval_interval │ 1 │
├──────────────────┼────────────────────────────┤
│ num_classes │ 8 │
├──────────────────┼────────────────────────────┤
│ depth │ 0.33 │
├──────────────────┼────────────────────────────┤
│ width │ 0.5 │
├──────────────────┼────────────────────────────┤
│ data_num_workers │ 0 │
├──────────────────┼────────────────────────────┤
│ input_size │ (640, 640) │
├──────────────────┼────────────────────────────┤
│ random_size │ (14, 26) │
├──────────────────┼────────────────────────────┤
│ data_dir │ 'datasets/coco128' │
├──────────────────┼────────────────────────────┤
│ train_ann │ 'instances_train2017.json' │
├──────────────────┼────────────────────────────┤
│ val_ann │ 'instances_val2017.json' │
├──────────────────┼────────────────────────────┤
│ degrees │ 10.0 │
├──────────────────┼────────────────────────────┤
│ translate │ 0.1 │
├──────────────────┼────────────────────────────┤
│ scale │ (0.1, 2) │
├──────────────────┼────────────────────────────┤
│ mscale │ (0.8, 1.6) │
├──────────────────┼────────────────────────────┤
│ shear │ 2.0 │
├──────────────────┼────────────────────────────┤
│ perspective │ 0.0 │
├──────────────────┼────────────────────────────┤
│ enable_mixup │ True │
├──────────────────┼────────────────────────────┤
│ warmup_epochs │ 5 │
├──────────────────┼────────────────────────────┤
│ max_epoch │ 10 │
├──────────────────┼────────────────────────────┤
│ warmup_lr │ 0 │
├──────────────────┼────────────────────────────┤
│ basic_lr_per_img │ 0.00015625 │
├──────────────────┼────────────────────────────┤
│ scheduler │ 'yoloxwarmcos' │
├──────────────────┼────────────────────────────┤
│ no_aug_epochs │ 15 │
├──────────────────┼────────────────────────────┤
│ min_lr_ratio │ 0.05 │
├──────────────────┼────────────────────────────┤
│ ema │ True │
├──────────────────┼────────────────────────────┤
│ weight_decay │ 0.0005 │
├──────────────────┼────────────────────────────┤
│ momentum │ 0.9 │
├──────────────────┼────────────────────────────┤
│ exp_name │ '606' │
├──────────────────┼────────────────────────────┤
│ test_size │ (640, 640) │
├──────────────────┼────────────────────────────┤
│ test_conf │ 0.01 │
├──────────────────┼────────────────────────────┤
│ nmsthre │ 0.65 │
╘══════════════════╧════════════════════════════╛
2021-08-15 21:54:11.239 | INFO | yolox.core.trainer:before_train:132 - Model Summary: Params: 8.94M, Gflops: 26.65
2021-08-15 21:54:13.208 | INFO | apex.amp.frontend:initialize:328 - Selected optimization level O1: Insert automatic casts around Pytorch functions and Tensor methods.
2021-08-15 21:54:13.208 | INFO | apex.amp.frontend:initialize:329 - Defaults for this optimization level are:
2021-08-15 21:54:13.209 | INFO | apex.amp.frontend:initialize:331 - enabled : True
2021-08-15 21:54:13.209 | INFO | apex.amp.frontend:initialize:331 - opt_level : O1
2021-08-15 21:54:13.209 | INFO | apex.amp.frontend:initialize:331 - cast_model_type : None
2021-08-15 21:54:13.210 | INFO | apex.amp.frontend:initialize:331 - patch_torch_functions : True
2021-08-15 21:54:13.210 | INFO | apex.amp.frontend:initialize:331 - keep_batchnorm_fp32 : None
2021-08-15 21:54:13.210 | INFO | apex.amp.frontend:initialize:331 - master_weights : None
2021-08-15 21:54:13.210 | INFO | apex.amp.frontend:initialize:331 - loss_scale : dynamic
2021-08-15 21:54:13.211 | INFO | apex.amp.frontend:initialize:336 - Processing user overrides (additional kwargs that are not None)...
2021-08-15 21:54:13.211 | INFO | apex.amp.frontend:initialize:354 - After processing overrides, optimization options are:
2021-08-15 21:54:13.211 | INFO | apex.amp.frontend:initialize:356 - enabled : True
2021-08-15 21:54:13.212 | INFO | apex.amp.frontend:initialize:356 - opt_level : O1
2021-08-15 21:54:13.213 | INFO | apex.amp.frontend:initialize:356 - cast_model_type : None
2021-08-15 21:54:13.214 | INFO | apex.amp.frontend:initialize:356 - patch_torch_functions : True
2021-08-15 21:54:13.215 | INFO | apex.amp.frontend:initialize:356 - keep_batchnorm_fp32 : None
2021-08-15 21:54:13.216 | INFO | apex.amp.frontend:initialize:356 - master_weights : None
2021-08-15 21:54:13.217 | INFO | apex.amp.frontend:initialize:356 - loss_scale : dynamic
2021-08-15 21:54:13.221 | INFO | apex.amp.scaler:init:64 - Warning: multi_tensor_applier fused unscale kernel is unavailable, possibly because apex was installed without --cuda_ext --cpp_ext. Using Python fallback. Original ImportError was: ModuleNotFoundError("No module named 'amp_C'")
2021-08-15 21:54:13.223 | INFO | yolox.core.trainer:resume_train:292 - loading checkpoint for fine tuning
2021-08-15 21:54:13.351 | WARNING | yolox.utils.checkpoint:load_ckpt:24 - Shape of head.cls_preds.0.weight in checkpoint is torch.Size([80, 128, 1, 1]), while shape of head.cls_preds.0.weight in model is torch.Size([8, 128, 1, 1]).
2021-08-15 21:54:13.352 | WARNING | yolox.utils.checkpoint:load_ckpt:24 - Shape of head.cls_preds.0.bias in checkpoint is torch.Size([80]), while shape of head.cls_preds.0.bias in model is torch.Size([8]).
2021-08-15 21:54:13.352 | WARNING | yolox.utils.checkpoint:load_ckpt:24 - Shape of head.cls_preds.1.weight in checkpoint is torch.Size([80, 128, 1, 1]), while shape of head.cls_preds.1.weight in model is torch.Size([8, 128, 1, 1]).
2021-08-15 21:54:13.352 | WARNING | yolox.utils.checkpoint:load_ckpt:24 - Shape of head.cls_preds.1.bias in checkpoint is torch.Size([80]), while shape of head.cls_preds.1.bias in model is torch.Size([8]).
2021-08-15 21:54:13.352 | WARNING | yolox.utils.checkpoint:load_ckpt:24 - Shape of head.cls_preds.2.weight in checkpoint is torch.Size([80, 128, 1, 1]), while shape of head.cls_preds.2.weight in model is torch.Size([8, 128, 1, 1]).
2021-08-15 21:54:13.353 | WARNING | yolox.utils.checkpoint:load_ckpt:24 - Shape of head.cls_preds.2.bias in checkpoint is torch.Size([80]), while shape of head.cls_preds.2.bias in model is torch.Size([8]).
2021-08-15 21:54:13.377 | INFO | yolox.data.datasets.coco:init:43 - loading annotations into memory...
2021-08-15 21:54:13.405 | INFO | yolox.data.datasets.coco:init:43 - Done (t=0.03s)
2021-08-15 21:54:13.406 | INFO | pycocotools.coco:init:89 - creating index...
2021-08-15 21:54:13.407 | INFO | pycocotools.coco:init:89 - index created!
2021-08-15 21:54:13.433 | INFO | yolox.core.trainer:before_train:153 - init prefetcher, this might take one minute or less...
2021-08-15 21:54:13.746 | INFO | yolox.data.datasets.coco:init:43 - loading annotations into memory...
2021-08-15 21:54:13.749 | INFO | yolox.data.datasets.coco:init:43 - Done (t=0.00s)
2021-08-15 21:54:13.749 | INFO | pycocotools.coco:init:89 - creating index...
2021-08-15 21:54:13.749 | INFO | pycocotools.coco:init:89 - index created!
2021-08-15 21:54:13.759 | INFO | yolox.core.trainer:before_train:183 - Training start...
2021-08-15 21:54:13.762 | INFO | yolox.core.trainer:before_train:184 -
YOLOX(
(backbone): YOLOPAFPN(
(backbone): CSPDarknet(
(stem): Focus(
(conv): BaseConv(
(conv): Conv2d(12, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
(dark2): Sequential(
(0): BaseConv(
(conv): Conv2d(32, 64, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): CSPLayer(
(conv1): BaseConv(
(conv): Conv2d(64, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv2): BaseConv(
(conv): Conv2d(64, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv3): BaseConv(
(conv): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): Sequential(
(0): Bottleneck(
(conv1): BaseConv(
(conv): Conv2d(32, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv2): BaseConv(
(conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
)
)
)
(dark3): Sequential(
(0): BaseConv(
(conv): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): CSPLayer(
(conv1): BaseConv(
(conv): Conv2d(128, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv2): BaseConv(
(conv): Conv2d(128, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv3): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): Sequential(
(0): Bottleneck(
(conv1): BaseConv(
(conv): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv2): BaseConv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
(1): Bottleneck(
(conv1): BaseConv(
(conv): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv2): BaseConv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
(2): Bottleneck(
(conv1): BaseConv(
(conv): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv2): BaseConv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
)
)
)
(dark4): Sequential(
(0): BaseConv(
(conv): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): CSPLayer(
(conv1): BaseConv(
(conv): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv2): BaseConv(
(conv): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv3): BaseConv(
(conv): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): Sequential(
(0): Bottleneck(
(conv1): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv2): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
(1): Bottleneck(
(conv1): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv2): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
(2): Bottleneck(
(conv1): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv2): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
)
)
)
(dark5): Sequential(
(0): BaseConv(
(conv): Conv2d(256, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): SPPBottleneck(
(conv1): BaseConv(
(conv): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): ModuleList(
(0): MaxPool2d(kernel_size=5, stride=1, padding=2, dilation=1, ceil_mode=False)
(1): MaxPool2d(kernel_size=9, stride=1, padding=4, dilation=1, ceil_mode=False)
(2): MaxPool2d(kernel_size=13, stride=1, padding=6, dilation=1, ceil_mode=False)
)
(conv2): BaseConv(
(conv): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
(2): CSPLayer(
(conv1): BaseConv(
(conv): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv2): BaseConv(
(conv): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv3): BaseConv(
(conv): Conv2d(512, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): Sequential(
(0): Bottleneck(
(conv1): BaseConv(
(conv): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv2): BaseConv(
(conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
)
)
)
)
(upsample): Upsample(scale_factor=2.0, mode=nearest)
(lateral_conv0): BaseConv(
(conv): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(C3_p4): CSPLayer(
(conv1): BaseConv(
(conv): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv2): BaseConv(
(conv): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv3): BaseConv(
(conv): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): Sequential(
(0): Bottleneck(
(conv1): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv2): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
)
)
(reduce_conv1): BaseConv(
(conv): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(C3_p3): CSPLayer(
(conv1): BaseConv(
(conv): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv2): BaseConv(
(conv): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv3): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): Sequential(
(0): Bottleneck(
(conv1): BaseConv(
(conv): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv2): BaseConv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
)
)
(bu_conv2): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(C3_n3): CSPLayer(
(conv1): BaseConv(
(conv): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv2): BaseConv(
(conv): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv3): BaseConv(
(conv): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): Sequential(
(0): Bottleneck(
(conv1): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv2): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
)
)
(bu_conv1): BaseConv(
(conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(C3_n4): CSPLayer(
(conv1): BaseConv(
(conv): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv2): BaseConv(
(conv): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv3): BaseConv(
(conv): Conv2d(512, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): Sequential(
(0): Bottleneck(
(conv1): BaseConv(
(conv): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(conv2): BaseConv(
(conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
)
)
)
(head): YOLOXHead(
(cls_convs): ModuleList(
(0): Sequential(
(0): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
(1): Sequential(
(0): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
(2): Sequential(
(0): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
)
(reg_convs): ModuleList(
(0): Sequential(
(0): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
(1): Sequential(
(0): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
(2): Sequential(
(0): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
)
(cls_preds): ModuleList(
(0): Conv2d(128, 8, kernel_size=(1, 1), stride=(1, 1))
(1): Conv2d(128, 8, kernel_size=(1, 1), stride=(1, 1))
(2): Conv2d(128, 8, kernel_size=(1, 1), stride=(1, 1))
)
(reg_preds): ModuleList(
(0): Conv2d(128, 4, kernel_size=(1, 1), stride=(1, 1))
(1): Conv2d(128, 4, kernel_size=(1, 1), stride=(1, 1))
(2): Conv2d(128, 4, kernel_size=(1, 1), stride=(1, 1))
)
(obj_preds): ModuleList(
(0): Conv2d(128, 1, kernel_size=(1, 1), stride=(1, 1))
(1): Conv2d(128, 1, kernel_size=(1, 1), stride=(1, 1))
(2): Conv2d(128, 1, kernel_size=(1, 1), stride=(1, 1))
)
(stems): ModuleList(
(0): BaseConv(
(conv): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): BaseConv(
(conv): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(2): BaseConv(
(conv): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
(l1_loss): L1Loss()
(bcewithlog_loss): BCEWithLogitsLoss()
(iou_loss): IOUloss()
)
)
2021-08-15 21:54:13.766 | INFO | yolox.core.trainer:before_epoch:192 - ---> start train epoch1
2021-08-15 21:54:13.766 | INFO | yolox.core.trainer:before_epoch:195 - --->No mosaic aug now!
2021-08-15 21:54:13.767 | INFO | yolox.core.trainer:before_epoch:197 - --->Add additional L1 loss now!
2021-08-15 21:54:16.905 | INFO | apex.amp.handle:skip_step:138 - Gradient overflow. Skipping step, loss scaler 0 reducing loss scale to 32768.0
2021-08-15 21:54:17.337 | INFO | apex.amp.handle:skip_step:138 - Gradient overflow. Skipping step, loss scaler 0 reducing loss scale to 16384.0
2021-08-15 21:54:17.771 | INFO | apex.amp.handle:skip_step:138 - Gradient overflow. Skipping step, loss scaler 0 reducing loss scale to 8192.0
2021-08-15 21:54:18.230 | INFO | apex.amp.handle:skip_step:138 - Gradient overflow. Skipping step, loss scaler 0 reducing loss scale to 4096.0
2021-08-15 21:54:18.684 | INFO | apex.amp.handle:skip_step:138 - Gradient overflow. Skipping step, loss scaler 0 reducing loss scale to 2048.0
2021-08-15 21:54:19.135 | INFO | apex.amp.handle:skip_step:138 - Gradient overflow. Skipping step, loss scaler 0 reducing loss scale to 1024.0
2021-08-15 21:54:20.592 | INFO | apex.amp.handle:skip_step:138 - Gradient overflow. Skipping step, loss scaler 0 reducing loss scale to 512.0
2021-08-15 21:54:21.105 | INFO | yolox.core.trainer:after_iter:242 - epoch: 1/10, iter: 10/88, mem: 3017Mb, iter_time: 0.733s, data_time: 0.208s, total_loss: 21.5, iou_loss: 4.2, l1_loss: 3.2, conf_loss: 12.5, cls_loss: 1.5, lr: 6.457e-07, size: 640, ETA: 0:10:38
2021-08-15 21:54:25.868 | INFO | yolox.core.trainer:after_iter:242 - epoch: 1/10, iter: 20/88, mem: 3017Mb, iter_time: 0.476s, data_time: 0.201s, total_loss: 18.7, iou_loss: 4.2, l1_loss: 2.9, conf_loss: 10.2, cls_loss: 1.5, lr: 2.583e-06, size: 640, ETA: 0:08:39
2021-08-15 21:54:33.831 | INFO | yolox.core.trainer:after_iter:242 - epoch: 1/10, iter: 30/88, mem: 3881Mb, iter_time: 0.796s, data_time: 0.254s, total_loss: 16.9, iou_loss: 4.1, l1_loss: 3.2, conf_loss: 8.0, cls_loss: 1.6, lr: 5.811e-06, size: 704, ETA: 0:09:28
2021-08-15 21:54:41.413 | INFO | yolox.core.trainer:after_iter:242 - epoch: 1/10, iter: 40/88, mem: 3881Mb, iter_time: 0.758s, data_time: 0.219s, total_loss: 14.7, iou_loss: 3.9, l1_loss: 2.7, conf_loss: 6.7, cls_loss: 1.3, lr: 1.033e-05, size: 672, ETA: 0:09:40
2021-08-15 21:54:46.458 | INFO | yolox.core.trainer:after_iter:242 - epoch: 1/10, iter: 50/88, mem: 3881Mb, iter_time: 0.504s, data_time: 0.219s, total_loss: 13.2, iou_loss: 3.5, l1_loss: 2.6, conf_loss: 6.0, cls_loss: 1.2, lr: 1.614e-05, size: 672, ETA: 0:09:02
2021-08-15 21:54:51.804 | INFO | yolox.core.trainer:after_iter:242 - epoch: 1/10, iter: 60/88, mem: 3881Mb, iter_time: 0.534s, data_time: 0.146s, total_loss: 12.2, iou_loss: 3.7, l1_loss: 2.3, conf_loss: 5.3, cls_loss: 0.9, lr: 2.324e-05, size: 512, ETA: 0:08:39
2021-08-15 21:54:55.538 | INFO | yolox.core.trainer:after_iter:242 - epoch: 1/10, iter: 70/88, mem: 3881Mb, iter_time: 0.373s, data_time: 0.143s, total_loss: 11.0, iou_loss: 3.5, l1_loss: 1.9, conf_loss: 4.8, cls_loss: 0.8, lr: 3.164e-05, size: 512, ETA: 0:08:02
2021-08-15 21:55:02.818 | INFO | yolox.core.trainer:after_iter:242 - epoch: 1/10, iter: 80/88, mem: 3881Mb, iter_time: 0.728s, data_time: 0.184s, total_loss: 8.8, iou_loss: 2.8, l1_loss: 1.4, conf_loss: 3.9, cls_loss: 0.7, lr: 4.132e-05, size: 576, ETA: 0:08:10
2021-08-15 21:55:10.084 | INFO | yolox.core.trainer:save_ckpt:324 - Save weights to ./YOLOX_outputs\606
2021-08-15 21:55:17.377 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:171 - Evaluate in main process...
2021-08-15 21:55:17.451 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:204 - Loading and preparing results...
2021-08-15 21:55:17.473 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:204 - DONE (t=0.02s)
2021-08-15 21:55:17.474 | INFO | pycocotools.coco:loadRes:365 - creating index...
2021-08-15 21:55:17.475 | INFO | pycocotools.coco:loadRes:365 - index created!
2021-08-15 21:55:17.546 | INFO | yolox.core.trainer:evaluate_and_save_model:315 -
Average forward time: 6.26 ms, Average NMS time: 1.11 ms, Average inference time: 7.37 ms
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000

2021-08-15 21:55:17.547 | INFO | yolox.core.trainer:save_ckpt:324 - Save weights to ./YOLOX_outputs\606
2021-08-15 21:55:17.696 | INFO | yolox.core.trainer:before_epoch:192 - ---> start train epoch2
2021-08-15 21:55:17.696 | INFO | yolox.core.trainer:before_epoch:195 - --->No mosaic aug now!
2021-08-15 21:55:17.697 | INFO | yolox.core.trainer:before_epoch:197 - --->Add additional L1 loss now!
2021-08-15 21:55:22.867 | INFO | yolox.core.trainer:after_iter:242 - epoch: 2/10, iter: 10/88, mem: 3881Mb, iter_time: 0.517s, data_time: 0.223s, total_loss: 7.6, iou_loss: 2.6, l1_loss: 1.3, conf_loss: 3.1, cls_loss: 0.7, lr: 6.201e-05, size: 640, ETA: 0:08:10
2021-08-15 21:55:26.814 | INFO | yolox.core.trainer:after_iter:242 - epoch: 2/10, iter: 20/88, mem: 3881Mb, iter_time: 0.394s, data_time: 0.143s, total_loss: 7.0, iou_loss: 2.6, l1_loss: 1.2, conf_loss: 2.5, cls_loss: 0.7, lr: 7.531e-05, size: 512, ETA: 0:07:47
2021-08-15 21:55:32.197 | INFO | yolox.core.trainer:after_iter:242 - epoch: 2/10, iter: 30/88, mem: 3881Mb, iter_time: 0.538s, data_time: 0.140s, total_loss: 5.2, iou_loss: 2.0, l1_loss: 0.8, conf_loss: 1.9, cls_loss: 0.6, lr: 8.990e-05, size: 544, ETA: 0:07:36
2021-08-15 21:55:40.492 | INFO | yolox.core.trainer:after_iter:242 - epoch: 2/10, iter: 40/88, mem: 3881Mb, iter_time: 0.829s, data_time: 0.271s, total_loss: 6.5, iou_loss: 2.1, l1_loss: 1.2, conf_loss: 2.6, cls_loss: 0.6, lr: 1.058e-04, size: 800, ETA: 0:07:44
2021-08-15 21:55:45.579 | INFO | yolox.core.trainer:after_iter:242 - epoch: 2/10, iter: 50/88, mem: 3881Mb, iter_time: 0.508s, data_time: 0.204s, total_loss: 5.5, iou_loss: 2.0, l1_loss: 0.9, conf_loss: 2.1, cls_loss: 0.5, lr: 1.230e-04, size: 640, ETA: 0:07:32
2021-08-15 21:55:53.270 | INFO | yolox.core.trainer:after_iter:242 - epoch: 2/10, iter: 60/88, mem: 3881Mb, iter_time: 0.769s, data_time: 0.252s, total_loss: 4.2, iou_loss: 1.6, l1_loss: 0.8, conf_loss: 1.5, cls_loss: 0.5, lr: 1.414e-04, size: 736, ETA: 0:07:34
2021-08-15 21:55:57.769 | INFO | yolox.core.trainer:after_iter:242 - epoch: 2/10, iter: 70/88, mem: 3881Mb, iter_time: 0.449s, data_time: 0.175s, total_loss: 5.2, iou_loss: 2.2, l1_loss: 0.9, conf_loss: 1.5, cls_loss: 0.6, lr: 1.612e-04, size: 544, ETA: 0:07:20
2021-08-15 21:56:01.607 | INFO | yolox.core.trainer:after_iter:242 - epoch: 2/10, iter: 80/88, mem: 3881Mb, iter_time: 0.383s, data_time: 0.136s, total_loss: 3.3, iou_loss: 1.3, l1_loss: 0.5, conf_loss: 1.0, cls_loss: 0.4, lr: 1.822e-04, size: 512, ETA: 0:07:04
2021-08-15 21:56:04.585 | INFO | yolox.core.trainer:save_ckpt:324 - Save weights to ./YOLOX_outputs\606
2021-08-15 21:56:10.799 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:171 - Evaluate in main process...
2021-08-15 21:56:10.809 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:204 - Loading and preparing results...
2021-08-15 21:56:10.818 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:204 - DONE (t=0.01s)
2021-08-15 21:56:10.819 | INFO | pycocotools.coco:loadRes:365 - creating index...
2021-08-15 21:56:10.819 | INFO | pycocotools.coco:loadRes:365 - index created!
2021-08-15 21:56:10.856 | INFO | yolox.core.trainer:evaluate_and_save_model:315 -
Average forward time: 6.30 ms, Average NMS time: 1.09 ms, Average inference time: 7.39 ms
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000

2021-08-15 21:56:10.856 | INFO | yolox.core.trainer:save_ckpt:324 - Save weights to ./YOLOX_outputs\606
2021-08-15 21:56:11.022 | INFO | yolox.core.trainer:before_epoch:192 - ---> start train epoch3
2021-08-15 21:56:11.023 | INFO | yolox.core.trainer:before_epoch:195 - --->No mosaic aug now!
2021-08-15 21:56:11.023 | INFO | yolox.core.trainer:before_epoch:197 - --->Add additional L1 loss now!
2021-08-15 21:56:16.385 | INFO | yolox.core.trainer:after_iter:242 - epoch: 3/10, iter: 10/88, mem: 3881Mb, iter_time: 0.536s, data_time: 0.226s, total_loss: 7.6, iou_loss: 2.2, l1_loss: 1.1, conf_loss: 3.6, cls_loss: 0.7, lr: 2.234e-04, size: 800, ETA: 0:06:44
2021-08-15 21:56:21.538 | INFO | yolox.core.trainer:after_iter:242 - epoch: 3/10, iter: 20/88, mem: 3881Mb, iter_time: 0.514s, data_time: 0.203s, total_loss: 5.3, iou_loss: 2.3, l1_loss: 1.0, conf_loss: 1.4, cls_loss: 0.6, lr: 2.480e-04, size: 576, ETA: 0:06:36
2021-08-15 21:56:26.631 | INFO | yolox.core.trainer:after_iter:242 - epoch: 3/10, iter: 30/88, mem: 3881Mb, iter_time: 0.509s, data_time: 0.218s, total_loss: 6.6, iou_loss: 2.5, l1_loss: 1.5, conf_loss: 1.9, cls_loss: 0.7, lr: 2.740e-04, size: 736, ETA: 0:06:28
2021-08-15 21:56:31.465 | INFO | yolox.core.trainer:after_iter:242 - epoch: 3/10, iter: 40/88, mem: 3881Mb, iter_time: 0.483s, data_time: 0.191s, total_loss: 5.9, iou_loss: 2.5, l1_loss: 1.0, conf_loss: 1.7, cls_loss: 0.8, lr: 3.012e-04, size: 544, ETA: 0:06:19
2021-08-15 21:56:35.555 | INFO | yolox.core.trainer:after_iter:242 - epoch: 3/10, iter: 50/88, mem: 3881Mb, iter_time: 0.408s, data_time: 0.150s, total_loss: 3.5, iou_loss: 1.5, l1_loss: 0.5, conf_loss: 1.1, cls_loss: 0.5, lr: 3.298e-04, size: 544, ETA: 0:06:09
2021-08-15 21:56:40.253 | INFO | yolox.core.trainer:after_iter:242 - epoch: 3/10, iter: 60/88, mem: 3881Mb, iter_time: 0.469s, data_time: 0.191s, total_loss: 4.6, iou_loss: 2.1, l1_loss: 1.0, conf_loss: 0.8, cls_loss: 0.7, lr: 3.596e-04, size: 640, ETA: 0:06:01
2021-08-15 21:56:44.927 | INFO | yolox.core.trainer:after_iter:242 - epoch: 3/10, iter: 70/88, mem: 3881Mb, iter_time: 0.467s, data_time: 0.189s, total_loss: 3.2, iou_loss: 1.6, l1_loss: 0.6, conf_loss: 0.4, cls_loss: 0.5, lr: 3.907e-04, size: 576, ETA: 0:05:53
2021-08-15 21:56:49.526 | INFO | yolox.core.trainer:after_iter:242 - epoch: 3/10, iter: 80/88, mem: 3881Mb, iter_time: 0.459s, data_time: 0.126s, total_loss: 9.3, iou_loss: 3.6, l1_loss: 2.1, conf_loss: 2.8, cls_loss: 0.8, lr: 4.231e-04, size: 448, ETA: 0:05:45
2021-08-15 21:56:52.664 | INFO | yolox.core.trainer:save_ckpt:324 - Save weights to ./YOLOX_outputs\606
2021-08-15 21:56:58.751 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:171 - Evaluate in main process...
2021-08-15 21:56:58.770 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:204 - Loading and preparing results...
2021-08-15 21:56:58.779 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:204 - DONE (t=0.01s)
2021-08-15 21:56:58.779 | INFO | pycocotools.coco:loadRes:365 - creating index...
2021-08-15 21:56:58.780 | INFO | pycocotools.coco:loadRes:365 - index created!
2021-08-15 21:56:58.851 | INFO | yolox.core.trainer:evaluate_and_save_model:315 -
Average forward time: 6.29 ms, Average NMS time: 1.10 ms, Average inference time: 7.39 ms
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000

2021-08-15 21:56:58.852 | INFO | yolox.core.trainer:save_ckpt:324 - Save weights to ./YOLOX_outputs\606
2021-08-15 21:56:59.009 | INFO | yolox.core.trainer:before_epoch:192 - ---> start train epoch4
2021-08-15 21:56:59.010 | INFO | yolox.core.trainer:before_epoch:195 - --->No mosaic aug now!
2021-08-15 21:56:59.010 | INFO | yolox.core.trainer:before_epoch:197 - --->Add additional L1 loss now!
2021-08-15 21:57:03.834 | INFO | yolox.core.trainer:after_iter:242 - epoch: 4/10, iter: 10/88, mem: 3881Mb, iter_time: 0.482s, data_time: 0.194s, total_loss: 3.5, iou_loss: 1.5, l1_loss: 0.6, conf_loss: 0.9, cls_loss: 0.5, lr: 4.847e-04, size: 640, ETA: 0:05:30
2021-08-15 21:57:08.106 | INFO | yolox.core.trainer:after_iter:242 - epoch: 4/10, iter: 20/88, mem: 3881Mb, iter_time: 0.427s, data_time: 0.159s, total_loss: 11.3, iou_loss: 4.4, l1_loss: 3.9, conf_loss: 2.2, cls_loss: 0.8, lr: 5.208e-04, size: 448, ETA: 0:05:22
2021-08-15 21:57:12.441 | INFO | yolox.core.trainer:after_iter:242 - epoch: 4/10, iter: 30/88, mem: 3881Mb, iter_time: 0.433s, data_time: 0.108s, total_loss: 5.2, iou_loss: 2.3, l1_loss: 0.9, conf_loss: 1.5, cls_loss: 0.6, lr: 5.581e-04, size: 480, ETA: 0:05:15
2021-08-15 21:57:16.038 | INFO | yolox.core.trainer:after_iter:242 - epoch: 4/10, iter: 40/88, mem: 3881Mb, iter_time: 0.359s, data_time: 0.125s, total_loss: 4.3, iou_loss: 2.0, l1_loss: 0.8, conf_loss: 0.8, cls_loss: 0.7, lr: 5.967e-04, size: 544, ETA: 0:05:06
2021-08-15 21:57:19.877 | INFO | yolox.core.trainer:after_iter:242 - epoch: 4/10, iter: 50/88, mem: 3881Mb, iter_time: 0.384s, data_time: 0.136s, total_loss: 2.7, iou_loss: 1.3, l1_loss: 0.4, conf_loss: 0.6, cls_loss: 0.4, lr: 6.366e-04, size: 512, ETA: 0:04:58
2021-08-15 21:57:24.603 | INFO | yolox.core.trainer:after_iter:242 - epoch: 4/10, iter: 60/88, mem: 3881Mb, iter_time: 0.472s, data_time: 0.196s, total_loss: 11.9, iou_loss: 3.5, l1_loss: 2.2, conf_loss: 5.3, cls_loss: 0.9, lr: 6.778e-04, size: 800, ETA: 0:04:52
2021-08-15 21:57:30.386 | INFO | yolox.core.trainer:after_iter:242 - epoch: 4/10, iter: 70/88, mem: 3881Mb, iter_time: 0.578s, data_time: 0.243s, total_loss: 7.8, iou_loss: 2.7, l1_loss: 1.4, conf_loss: 3.1, cls_loss: 0.7, lr: 7.203e-04, size: 576, ETA: 0:04:47
2021-08-15 21:57:34.968 | INFO | yolox.core.trainer:after_iter:242 - epoch: 4/10, iter: 80/88, mem: 3881Mb, iter_time: 0.457s, data_time: 0.187s, total_loss: 5.8, iou_loss: 2.3, l1_loss: 1.2, conf_loss: 1.8, cls_loss: 0.6, lr: 7.640e-04, size: 704, ETA: 0:04:41
2021-08-15 21:57:38.888 | INFO | yolox.core.trainer:save_ckpt:324 - Save weights to ./YOLOX_outputs\606
2021-08-15 21:57:45.180 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:171 - Evaluate in main process...
2021-08-15 21:57:45.223 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:204 - Loading and preparing results...
2021-08-15 21:57:45.241 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:204 - DONE (t=0.02s)
2021-08-15 21:57:45.241 | INFO | pycocotools.coco:loadRes:365 - creating index...
2021-08-15 21:57:45.242 | INFO | pycocotools.coco:loadRes:365 - index created!
2021-08-15 21:57:45.331 | INFO | yolox.core.trainer:evaluate_and_save_model:315 -
Average forward time: 6.29 ms, Average NMS time: 1.14 ms, Average inference time: 7.43 ms
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000

2021-08-15 21:57:45.332 | INFO | yolox.core.trainer:save_ckpt:324 - Save weights to ./YOLOX_outputs\606
2021-08-15 21:57:45.498 | INFO | yolox.core.trainer:before_epoch:192 - ---> start train epoch5
2021-08-15 21:57:45.498 | INFO | yolox.core.trainer:before_epoch:195 - --->No mosaic aug now!
2021-08-15 21:57:45.498 | INFO | yolox.core.trainer:before_epoch:197 - --->Add additional L1 loss now!
2021-08-15 21:57:50.643 | INFO | yolox.core.trainer:after_iter:242 - epoch: 5/10, iter: 10/88, mem: 3881Mb, iter_time: 0.513s, data_time: 0.117s, total_loss: 12.8, iou_loss: 3.9, l1_loss: 2.5, conf_loss: 5.6, cls_loss: 0.8, lr: 8.461e-04, size: 608, ETA: 0:04:31
2021-08-15 21:57:58.209 | INFO | yolox.core.trainer:after_iter:242 - epoch: 5/10, iter: 20/88, mem: 3881Mb, iter_time: 0.756s, data_time: 0.216s, total_loss: 16.1, iou_loss: 4.5, l1_loss: 4.9, conf_loss: 5.5, cls_loss: 1.2, lr: 8.935e-04, size: 832, ETA: 0:04:29
2021-08-15 21:58:04.978 | INFO | yolox.core.trainer:after_iter:242 - epoch: 5/10, iter: 30/88, mem: 3881Mb, iter_time: 0.676s, data_time: 0.296s, total_loss: 20.1, iou_loss: 4.8, l1_loss: 5.6, conf_loss: 8.5, cls_loss: 1.2, lr: 9.422e-04, size: 448, ETA: 0:04:25
2021-08-15 21:58:08.970 | INFO | yolox.core.trainer:after_iter:242 - epoch: 5/10, iter: 40/88, mem: 3881Mb, iter_time: 0.398s, data_time: 0.132s, total_loss: 8.5, iou_loss: 3.6, l1_loss: 2.0, conf_loss: 2.1, cls_loss: 0.8, lr: 9.921e-04, size: 544, ETA: 0:04:18
2021-08-15 21:58:13.747 | INFO | yolox.core.trainer:after_iter:242 - epoch: 5/10, iter: 50/88, mem: 3881Mb, iter_time: 0.477s, data_time: 0.198s, total_loss: 18.3, iou_loss: 4.0, l1_loss: 2.4, conf_loss: 10.6, cls_loss: 1.3, lr: 1.043e-03, size: 832, ETA: 0:04:12
2021-08-15 21:58:20.429 | INFO | yolox.core.trainer:after_iter:242 - epoch: 5/10, iter: 60/88, mem: 3881Mb, iter_time: 0.667s, data_time: 0.292s, total_loss: 11.5, iou_loss: 4.4, l1_loss: 2.9, conf_loss: 3.1, cls_loss: 1.1, lr: 1.096e-03, size: 544, ETA: 0:04:09
2021-08-15 21:58:24.532 | INFO | yolox.core.trainer:after_iter:242 - epoch: 5/10, iter: 70/88, mem: 3881Mb, iter_time: 0.410s, data_time: 0.155s, total_loss: 7.2, iou_loss: 3.2, l1_loss: 1.0, conf_loss: 2.1, cls_loss: 0.7, lr: 1.150e-03, size: 608, ETA: 0:04:02
2021-08-15 21:58:28.869 | INFO | yolox.core.trainer:after_iter:242 - epoch: 5/10, iter: 80/88, mem: 3881Mb, iter_time: 0.433s, data_time: 0.164s, total_loss: 11.0, iou_loss: 4.2, l1_loss: 1.8, conf_loss: 4.1, cls_loss: 1.0, lr: 1.205e-03, size: 480, ETA: 0:03:56
2021-08-15 21:58:31.620 | INFO | yolox.core.trainer:save_ckpt:324 - Save weights to ./YOLOX_outputs\606
2021-08-15 21:58:38.010 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:171 - Evaluate in main process...
2021-08-15 21:58:38.093 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:204 - Loading and preparing results...
2021-08-15 21:58:38.158 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:204 - DONE (t=0.06s)
2021-08-15 21:58:38.158 | INFO | pycocotools.coco:loadRes:365 - creating index...
2021-08-15 21:58:38.160 | INFO | pycocotools.coco:loadRes:365 - index created!
2021-08-15 21:58:38.224 | INFO | yolox.core.trainer:evaluate_and_save_model:315 -
Average forward time: 6.54 ms, Average NMS time: 1.14 ms, Average inference time: 7.68 ms
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000

2021-08-15 21:58:38.225 | INFO | yolox.core.trainer:save_ckpt:324 - Save weights to ./YOLOX_outputs\606
2021-08-15 21:58:38.376 | INFO | yolox.core.trainer:before_epoch:192 - ---> start train epoch6
2021-08-15 21:58:38.376 | INFO | yolox.core.trainer:before_epoch:195 - --->No mosaic aug now!
2021-08-15 21:58:38.377 | INFO | yolox.core.trainer:before_epoch:197 - --->Add additional L1 loss now!
2021-08-15 21:58:44.452 | INFO | yolox.core.trainer:after_iter:242 - epoch: 6/10, iter: 10/88, mem: 3881Mb, iter_time: 0.607s, data_time: 0.277s, total_loss: 10.6, iou_loss: 4.4, l1_loss: 2.5, conf_loss: 3.0, cls_loss: 0.7, lr: 6.250e-05, size: 768, ETA: 0:03:46
2021-08-15 21:58:47.853 | INFO | yolox.core.trainer:after_iter:242 - epoch: 6/10, iter: 20/88, mem: 3881Mb, iter_time: 0.340s, data_time: 0.101s, total_loss: 5.3, iou_loss: 2.6, l1_loss: 0.6, conf_loss: 1.4, cls_loss: 0.7, lr: 6.250e-05, size: 448, ETA: 0:03:39
2021-08-15 21:58:54.795 | INFO | yolox.core.trainer:after_iter:242 - epoch: 6/10, iter: 30/88, mem: 3881Mb, iter_time: 0.693s, data_time: 0.325s, total_loss: 12.1, iou_loss: 4.3, l1_loss: 2.6, conf_loss: 4.5, cls_loss: 0.7, lr: 6.250e-05, size: 832, ETA: 0:03:35
2021-08-15 21:58:58.460 | INFO | yolox.core.trainer:after_iter:242 - epoch: 6/10, iter: 40/88, mem: 3881Mb, iter_time: 0.366s, data_time: 0.117s, total_loss: 5.1, iou_loss: 2.0, l1_loss: 0.4, conf_loss: 2.1, cls_loss: 0.6, lr: 6.250e-05, size: 480, ETA: 0:03:28
2021-08-15 21:59:04.487 | INFO | yolox.core.trainer:after_iter:242 - epoch: 6/10, iter: 50/88, mem: 3881Mb, iter_time: 0.602s, data_time: 0.278s, total_loss: 9.1, iou_loss: 3.9, l1_loss: 1.8, conf_loss: 2.7, cls_loss: 0.7, lr: 6.250e-05, size: 768, ETA: 0:03:24
2021-08-15 21:59:10.483 | INFO | yolox.core.trainer:after_iter:242 - epoch: 6/10, iter: 60/88, mem: 3881Mb, iter_time: 0.599s, data_time: 0.269s, total_loss: 8.2, iou_loss: 3.2, l1_loss: 1.3, conf_loss: 3.0, cls_loss: 0.7, lr: 6.250e-05, size: 768, ETA: 0:03:19
2021-08-15 21:59:15.368 | INFO | yolox.core.trainer:after_iter:242 - epoch: 6/10, iter: 70/88, mem: 3881Mb, iter_time: 0.488s, data_time: 0.196s, total_loss: 4.9, iou_loss: 2.3, l1_loss: 0.6, conf_loss: 1.4, cls_loss: 0.6, lr: 6.250e-05, size: 640, ETA: 0:03:14
2021-08-15 21:59:21.506 | INFO | yolox.core.trainer:after_iter:242 - epoch: 6/10, iter: 80/88, mem: 3881Mb, iter_time: 0.613s, data_time: 0.283s, total_loss: 5.8, iou_loss: 2.4, l1_loss: 0.8, conf_loss: 2.0, cls_loss: 0.6, lr: 6.250e-05, size: 736, ETA: 0:03:09
2021-08-15 21:59:26.364 | INFO | yolox.core.trainer:save_ckpt:324 - Save weights to ./YOLOX_outputs\606
2021-08-15 21:59:33.032 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:171 - Evaluate in main process...
2021-08-15 21:59:33.053 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:204 - Loading and preparing results...
2021-08-15 21:59:33.080 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:204 - DONE (t=0.03s)
2021-08-15 21:59:33.080 | INFO | pycocotools.coco:loadRes:365 - creating index...
2021-08-15 21:59:33.081 | INFO | pycocotools.coco:loadRes:365 - index created!
2021-08-15 21:59:33.134 | INFO | yolox.core.trainer:evaluate_and_save_model:315 -
Average forward time: 6.32 ms, Average NMS time: 1.17 ms, Average inference time: 7.49 ms
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000

2021-08-15 21:59:33.134 | INFO | yolox.core.trainer:save_ckpt:324 - Save weights to ./YOLOX_outputs\606
2021-08-15 21:59:33.312 | INFO | yolox.core.trainer:before_epoch:192 - ---> start train epoch7
2021-08-15 21:59:33.313 | INFO | yolox.core.trainer:before_epoch:195 - --->No mosaic aug now!
2021-08-15 21:59:33.313 | INFO | yolox.core.trainer:before_epoch:197 - --->Add additional L1 loss now!
2021-08-15 21:59:38.130 | INFO | yolox.core.trainer:after_iter:242 - epoch: 7/10, iter: 10/88, mem: 3881Mb, iter_time: 0.481s, data_time: 0.192s, total_loss: 6.9, iou_loss: 3.5, l1_loss: 1.4, conf_loss: 1.3, cls_loss: 0.8, lr: 6.250e-05, size: 576, ETA: 0:03:00
2021-08-15 21:59:43.472 | INFO | yolox.core.trainer:after_iter:242 - epoch: 7/10, iter: 20/88, mem: 3881Mb, iter_time: 0.533s, data_time: 0.231s, total_loss: 4.0, iou_loss: 1.7, l1_loss: 0.5, conf_loss: 1.2, cls_loss: 0.6, lr: 6.250e-05, size: 704, ETA: 0:02:54
2021-08-15 21:59:48.087 | INFO | yolox.core.trainer:after_iter:242 - epoch: 7/10, iter: 30/88, mem: 3881Mb, iter_time: 0.461s, data_time: 0.181s, total_loss: 4.7, iou_loss: 2.4, l1_loss: 0.7, conf_loss: 1.1, cls_loss: 0.6, lr: 6.250e-05, size: 576, ETA: 0:02:49
2021-08-15 21:59:53.363 | INFO | yolox.core.trainer:after_iter:242 - epoch: 7/10, iter: 40/88, mem: 3881Mb, iter_time: 0.527s, data_time: 0.228s, total_loss: 5.8, iou_loss: 2.7, l1_loss: 0.9, conf_loss: 1.6, cls_loss: 0.6, lr: 6.250e-05, size: 736, ETA: 0:02:44
2021-08-15 21:59:58.929 | INFO | yolox.core.trainer:after_iter:242 - epoch: 7/10, iter: 50/88, mem: 3881Mb, iter_time: 0.556s, data_time: 0.244s, total_loss: 4.1, iou_loss: 1.7, l1_loss: 0.5, conf_loss: 1.3, cls_loss: 0.5, lr: 6.250e-05, size: 736, ETA: 0:02:38
2021-08-15 22:00:05.210 | INFO | yolox.core.trainer:after_iter:242 - epoch: 7/10, iter: 60/88, mem: 3881Mb, iter_time: 0.628s, data_time: 0.278s, total_loss: 3.8, iou_loss: 1.6, l1_loss: 0.5, conf_loss: 1.3, cls_loss: 0.5, lr: 6.250e-05, size: 800, ETA: 0:02:34
2021-08-15 22:00:09.728 | INFO | yolox.core.trainer:after_iter:242 - epoch: 7/10, iter: 70/88, mem: 3881Mb, iter_time: 0.451s, data_time: 0.167s, total_loss: 8.5, iou_loss: 3.8, l1_loss: 1.8, conf_loss: 2.0, cls_loss: 0.8, lr: 6.250e-05, size: 544, ETA: 0:02:28
2021-08-15 22:00:14.975 | INFO | yolox.core.trainer:after_iter:242 - epoch: 7/10, iter: 80/88, mem: 3881Mb, iter_time: 0.524s, data_time: 0.225s, total_loss: 3.0, iou_loss: 1.1, l1_loss: 0.3, conf_loss: 1.0, cls_loss: 0.6, lr: 6.250e-05, size: 736, ETA: 0:02:23
2021-08-15 22:00:19.952 | INFO | yolox.core.trainer:save_ckpt:324 - Save weights to ./YOLOX_outputs\606
2021-08-15 22:00:25.995 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:171 - Evaluate in main process...
2021-08-15 22:00:26.010 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:204 - Loading and preparing results...
2021-08-15 22:00:26.018 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:204 - DONE (t=0.01s)
2021-08-15 22:00:26.018 | INFO | pycocotools.coco:loadRes:365 - creating index...
2021-08-15 22:00:26.019 | INFO | pycocotools.coco:loadRes:365 - index created!
2021-08-15 22:00:26.058 | INFO | yolox.core.trainer:evaluate_and_save_model:315 -
Average forward time: 6.28 ms, Average NMS time: 1.08 ms, Average inference time: 7.36 ms
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000

2021-08-15 22:00:26.058 | INFO | yolox.core.trainer:save_ckpt:324 - Save weights to ./YOLOX_outputs\606
2021-08-15 22:00:26.221 | INFO | yolox.core.trainer:before_epoch:192 - ---> start train epoch8
2021-08-15 22:00:26.221 | INFO | yolox.core.trainer:before_epoch:195 - --->No mosaic aug now!
2021-08-15 22:00:26.222 | INFO | yolox.core.trainer:before_epoch:197 - --->Add additional L1 loss now!
2021-08-15 22:00:31.045 | INFO | yolox.core.trainer:after_iter:242 - epoch: 8/10, iter: 10/88, mem: 3881Mb, iter_time: 0.482s, data_time: 0.183s, total_loss: 10.0, iou_loss: 3.8, l1_loss: 2.1, conf_loss: 3.2, cls_loss: 0.9, lr: 6.250e-05, size: 512, ETA: 0:02:13
2021-08-15 22:00:35.180 | INFO | yolox.core.trainer:after_iter:242 - epoch: 8/10, iter: 20/88, mem: 3881Mb, iter_time: 0.413s, data_time: 0.155s, total_loss: 5.6, iou_loss: 2.6, l1_loss: 1.0, conf_loss: 1.4, cls_loss: 0.7, lr: 6.250e-05, size: 640, ETA: 0:02:08
2021-08-15 22:00:39.925 | INFO | yolox.core.trainer:after_iter:242 - epoch: 8/10, iter: 30/88, mem: 3881Mb, iter_time: 0.474s, data_time: 0.191s, total_loss: 3.9, iou_loss: 1.8, l1_loss: 0.5, conf_loss: 1.0, cls_loss: 0.6, lr: 6.250e-05, size: 672, ETA: 0:02:02
2021-08-15 22:00:44.020 | INFO | yolox.core.trainer:after_iter:242 - epoch: 8/10, iter: 40/88, mem: 3881Mb, iter_time: 0.409s, data_time: 0.150s, total_loss: 6.4, iou_loss: 3.3, l1_loss: 1.2, conf_loss: 1.1, cls_loss: 0.8, lr: 6.250e-05, size: 512, ETA: 0:01:57
2021-08-15 22:00:48.270 | INFO | yolox.core.trainer:after_iter:242 - epoch: 8/10, iter: 50/88, mem: 3881Mb, iter_time: 0.424s, data_time: 0.167s, total_loss: 4.1, iou_loss: 2.0, l1_loss: 0.6, conf_loss: 1.0, cls_loss: 0.6, lr: 6.250e-05, size: 672, ETA: 0:01:51
2021-08-15 22:00:52.262 | INFO | yolox.core.trainer:after_iter:242 - epoch: 8/10, iter: 60/88, mem: 3881Mb, iter_time: 0.398s, data_time: 0.140s, total_loss: 8.9, iou_loss: 3.9, l1_loss: 1.8, conf_loss: 2.4, cls_loss: 0.9, lr: 6.250e-05, size: 448, ETA: 0:01:45
2021-08-15 22:00:57.304 | INFO | yolox.core.trainer:after_iter:242 - epoch: 8/10, iter: 70/88, mem: 3881Mb, iter_time: 0.504s, data_time: 0.216s, total_loss: 5.0, iou_loss: 2.3, l1_loss: 0.8, conf_loss: 1.3, cls_loss: 0.7, lr: 6.250e-05, size: 768, ETA: 0:01:40
2021-08-15 22:01:03.148 | INFO | yolox.core.trainer:after_iter:242 - epoch: 8/10, iter: 80/88, mem: 3881Mb, iter_time: 0.584s, data_time: 0.259s, total_loss: 2.8, iou_loss: 1.1, l1_loss: 0.3, conf_loss: 0.9, cls_loss: 0.5, lr: 6.250e-05, size: 736, ETA: 0:01:35
2021-08-15 22:01:07.368 | INFO | yolox.core.trainer:save_ckpt:324 - Save weights to ./YOLOX_outputs\606
2021-08-15 22:01:13.748 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:171 - Evaluate in main process...
2021-08-15 22:01:13.761 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:204 - Loading and preparing results...
2021-08-15 22:01:13.771 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:204 - DONE (t=0.01s)
2021-08-15 22:01:13.771 | INFO | pycocotools.coco:loadRes:365 - creating index...
2021-08-15 22:01:13.772 | INFO | pycocotools.coco:loadRes:365 - index created!
2021-08-15 22:01:13.809 | INFO | yolox.core.trainer:evaluate_and_save_model:315 -
Average forward time: 6.27 ms, Average NMS time: 1.13 ms, Average inference time: 7.40 ms
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000

2021-08-15 22:01:13.810 | INFO | yolox.core.trainer:save_ckpt:324 - Save weights to ./YOLOX_outputs\606
2021-08-15 22:01:13.969 | INFO | yolox.core.trainer:before_epoch:192 - ---> start train epoch9
2021-08-15 22:01:13.969 | INFO | yolox.core.trainer:before_epoch:195 - --->No mosaic aug now!
2021-08-15 22:01:13.969 | INFO | yolox.core.trainer:before_epoch:197 - --->Add additional L1 loss now!
2021-08-15 22:01:19.724 | INFO | yolox.core.trainer:after_iter:242 - epoch: 9/10, iter: 10/88, mem: 3881Mb, iter_time: 0.575s, data_time: 0.261s, total_loss: 5.1, iou_loss: 2.2, l1_loss: 0.7, conf_loss: 1.5, cls_loss: 0.6, lr: 6.250e-05, size: 768, ETA: 0:01:26
2021-08-15 22:01:25.460 | INFO | yolox.core.trainer:after_iter:242 - epoch: 9/10, iter: 20/88, mem: 3881Mb, iter_time: 0.573s, data_time: 0.249s, total_loss: 5.0, iou_loss: 2.4, l1_loss: 0.8, conf_loss: 1.1, cls_loss: 0.7, lr: 6.250e-05, size: 640, ETA: 0:01:21
2021-08-15 22:01:30.151 | INFO | yolox.core.trainer:after_iter:242 - epoch: 9/10, iter: 30/88, mem: 3881Mb, iter_time: 0.469s, data_time: 0.185s, total_loss: 4.9, iou_loss: 2.4, l1_loss: 0.7, conf_loss: 1.1, cls_loss: 0.6, lr: 6.250e-05, size: 544, ETA: 0:01:16
2021-08-15 22:01:33.860 | INFO | yolox.core.trainer:after_iter:242 - epoch: 9/10, iter: 40/88, mem: 3881Mb, iter_time: 0.370s, data_time: 0.127s, total_loss: 5.9, iou_loss: 2.6, l1_loss: 0.7, conf_loss: 2.0, cls_loss: 0.6, lr: 6.250e-05, size: 480, ETA: 0:01:10
2021-08-15 22:01:38.131 | INFO | yolox.core.trainer:after_iter:242 - epoch: 9/10, iter: 50/88, mem: 3881Mb, iter_time: 0.427s, data_time: 0.167s, total_loss: 8.6, iou_loss: 3.4, l1_loss: 1.6, conf_loss: 2.9, cls_loss: 0.8, lr: 6.250e-05, size: 768, ETA: 0:01:05
2021-08-15 22:01:43.483 | INFO | yolox.core.trainer:after_iter:242 - epoch: 9/10, iter: 60/88, mem: 3881Mb, iter_time: 0.534s, data_time: 0.224s, total_loss: 4.4, iou_loss: 2.3, l1_loss: 0.5, conf_loss: 0.9, cls_loss: 0.6, lr: 6.250e-05, size: 576, ETA: 0:01:00
2021-08-15 22:01:48.007 | INFO | yolox.core.trainer:after_iter:242 - epoch: 9/10, iter: 70/88, mem: 3881Mb, iter_time: 0.452s, data_time: 0.178s, total_loss: 3.8, iou_loss: 1.8, l1_loss: 0.5, conf_loss: 1.0, cls_loss: 0.6, lr: 6.250e-05, size: 672, ETA: 0:00:54
2021-08-15 22:01:52.385 | INFO | yolox.core.trainer:after_iter:242 - epoch: 9/10, iter: 80/88, mem: 3881Mb, iter_time: 0.437s, data_time: 0.165s, total_loss: 8.9, iou_loss: 3.7, l1_loss: 1.6, conf_loss: 2.7, cls_loss: 1.0, lr: 6.250e-05, size: 448, ETA: 0:00:49
2021-08-15 22:01:55.374 | INFO | yolox.core.trainer:save_ckpt:324 - Save weights to ./YOLOX_outputs\606
2021-08-15 22:02:01.640 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:171 - Evaluate in main process...
2021-08-15 22:02:01.653 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:204 - Loading and preparing results...
2021-08-15 22:02:01.660 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:204 - DONE (t=0.01s)
2021-08-15 22:02:01.661 | INFO | pycocotools.coco:loadRes:365 - creating index...
2021-08-15 22:02:01.661 | INFO | pycocotools.coco:loadRes:365 - index created!
2021-08-15 22:02:01.701 | INFO | yolox.core.trainer:evaluate_and_save_model:315 -
Average forward time: 6.29 ms, Average NMS time: 1.07 ms, Average inference time: 7.36 ms
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000

2021-08-15 22:02:01.701 | INFO | yolox.core.trainer:save_ckpt:324 - Save weights to ./YOLOX_outputs\606
2021-08-15 22:02:01.865 | INFO | yolox.core.trainer:before_epoch:192 - ---> start train epoch10
2021-08-15 22:02:01.866 | INFO | yolox.core.trainer:before_epoch:195 - --->No mosaic aug now!
2021-08-15 22:02:01.866 | INFO | yolox.core.trainer:before_epoch:197 - --->Add additional L1 loss now!
2021-08-15 22:02:06.901 | INFO | yolox.core.trainer:after_iter:242 - epoch: 10/10, iter: 10/88, mem: 3881Mb, iter_time: 0.503s, data_time: 0.206s, total_loss: 3.8, iou_loss: 1.9, l1_loss: 0.5, conf_loss: 0.8, cls_loss: 0.5, lr: 6.250e-05, size: 608, ETA: 0:00:40
2021-08-15 22:02:11.762 | INFO | yolox.core.trainer:after_iter:242 - epoch: 10/10, iter: 20/88, mem: 3881Mb, iter_time: 0.486s, data_time: 0.204s, total_loss: 6.7, iou_loss: 3.0, l1_loss: 1.2, conf_loss: 1.7, cls_loss: 0.8, lr: 6.250e-05, size: 768, ETA: 0:00:34
2021-08-15 22:02:17.996 | INFO | yolox.core.trainer:after_iter:242 - epoch: 10/10, iter: 30/88, mem: 3881Mb, iter_time: 0.623s, data_time: 0.285s, total_loss: 7.2, iou_loss: 3.4, l1_loss: 1.5, conf_loss: 1.5, cls_loss: 0.8, lr: 6.250e-05, size: 832, ETA: 0:00:29
2021-08-15 22:02:24.433 | INFO | yolox.core.trainer:after_iter:242 - epoch: 10/10, iter: 40/88, mem: 3881Mb, iter_time: 0.643s, data_time: 0.276s, total_loss: 6.7, iou_loss: 3.4, l1_loss: 1.4, conf_loss: 1.2, cls_loss: 0.7, lr: 6.250e-05, size: 576, ETA: 0:00:24
2021-08-15 22:02:28.987 | INFO | yolox.core.trainer:after_iter:242 - epoch: 10/10, iter: 50/88, mem: 3881Mb, iter_time: 0.455s, data_time: 0.186s, total_loss: 4.9, iou_loss: 2.3, l1_loss: 0.8, conf_loss: 1.1, cls_loss: 0.6, lr: 6.250e-05, size: 832, ETA: 0:00:19
2021-08-15 22:02:35.567 | INFO | yolox.core.trainer:after_iter:242 - epoch: 10/10, iter: 60/88, mem: 3881Mb, iter_time: 0.658s, data_time: 0.286s, total_loss: 4.9, iou_loss: 2.4, l1_loss: 0.8, conf_loss: 1.1, cls_loss: 0.6, lr: 6.250e-05, size: 640, ETA: 0:00:14
2021-08-15 22:02:40.259 | INFO | yolox.core.trainer:after_iter:242 - epoch: 10/10, iter: 70/88, mem: 3881Mb, iter_time: 0.469s, data_time: 0.189s, total_loss: 3.3, iou_loss: 1.5, l1_loss: 0.4, conf_loss: 0.9, cls_loss: 0.5, lr: 6.250e-05, size: 640, ETA: 0:00:09
2021-08-15 22:02:45.042 | INFO | yolox.core.trainer:after_iter:242 - epoch: 10/10, iter: 80/88, mem: 3881Mb, iter_time: 0.478s, data_time: 0.196s, total_loss: 3.7, iou_loss: 1.9, l1_loss: 0.6, conf_loss: 0.6, cls_loss: 0.5, lr: 6.250e-05, size: 704, ETA: 0:00:04
2021-08-15 22:02:49.305 | INFO | yolox.core.trainer:save_ckpt:324 - Save weights to ./YOLOX_outputs\606
2021-08-15 22:02:55.436 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:171 - Evaluate in main process...
2021-08-15 22:02:55.449 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:204 - Loading and preparing results...
2021-08-15 22:02:55.457 | INFO | yolox.evaluators.coco_evaluator:evaluate_prediction:204 - DONE (t=0.01s)
2021-08-15 22:02:55.457 | INFO | pycocotools.coco:loadRes:365 - creating index...
2021-08-15 22:02:55.458 | INFO | pycocotools.coco:loadRes:365 - index created!
2021-08-15 22:02:55.498 | INFO | yolox.core.trainer:evaluate_and_save_model:315 -
Average forward time: 6.26 ms, Average NMS time: 1.10 ms, Average inference time: 7.36 ms
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000

2021-08-15 22:02:55.499 | INFO | yolox.core.trainer:save_ckpt:324 - Save weights to ./YOLOX_outputs\606
2021-08-15 22:02:55.667 | INFO | yolox.core.trainer:after_train:187 - Training of experiment is done and the best AP is 0.00

As you can see I only train for 10 epochs just for testing but the mAP is 0.00 and I get the same result even when I train for 300 epochs,and this is my own Exp file:
`#!/usr/bin/env python3

-- coding:utf-8 --

Copyright (c) Megvii, Inc. and its affiliates.

import os

from yolox.exp import Exp as MyExp

class Exp(MyExp):
def init(self):
super(Exp, self).init()
self.depth = 0.33
self.width = 0.50
self.exp_name = os.path.split(os.path.realpath(file))[1].split(".")[0]

    # Define yourself dataset path
    self.data_dir = "datasets/coco128"
    self.train_ann = "instances_train2017.json"
    self.val_ann = "instances_val2017.json"

    self.num_classes = 8

    self.max_epoch = 10
    self.data_num_workers = 0
    self.eval_interval = 1`

and the coco_classes.py:
`#!/usr/bin/env python3

-- coding:utf-8 --

Copyright (c) Megvii, Inc. and its affiliates.

COCO_CLASSES = (
"blue1",
"red1",
"blue2",
"red2",
"blue3",
"red3",
"blue4",
"red4",
)`
So what's wrong with my code?I have struggled into this a whole day,Honestly,I'm about to be crazy now!!!

The text was updated successfully, but these errors were encountered:

Bachelorwangwei · 2021-08-16T01:23:08Z

我也遇到了，loss是正常下降的，但是ap就是0

Hezhexi2002 · 2021-08-16T02:22:32Z

我也遇到了，loss是正常下降的，但是ap就是0

所以你解决了吗？你也是coco格式数据集训练的吗？

EdisionWew · 2021-08-16T02:57:21Z

我也遇到了，训练过程评估都是0，我是自己的数据转成的 coco格式

Hezhexi2002 · 2021-08-16T06:52:02Z

我现在打算用train_custom.md里的那个mini-coco128试一试，但是那是在google - drive上的我又下不下来:-(，不过我看了一下这个mini-coco128的文件结构和我转化后的coco格式数据集是一样的，真的不知道是怎么回事，而且我昨天还用那个训练后的权重试了一下视频推理，平均每帧需要0.3s，我是用的yolox-s，我用yolov5s的话每帧平均7ms，这让我也很吃惊

Hezhexi2002 · 2021-08-17T01:02:32Z

I have met same problem, but my datasets are in VOC format.

2021-08-16 19:13:27 | INFO     | yolox.evaluators.voc_evaluator:160 - Evaluate in main process...
Writing stain VOC results file
Writing flaw VOC results file
Writing burn VOC results file
Eval IoU : 0.50
AP for stain = 0.0000
AP for flaw = 0.0000
AP for burn = 0.0000
Mean AP = 0.0000
~~~~~~~~
Results:
0.000
0.000
0.000
0.000
~~~~~~~~

2021-08-16 19:13:27 | INFO     | yolox.core.trainer:321 - 
Average forward time: 0.00 ms, Average NMS time: 0.00 ms, Average inference time: 0.00 ms

顺便问一下你解决了这个问题吗？我的数据在YOLO5上是可以训练的

还没有：-（，但是voc格式mAP为0已经有解决方案了，你在issue可以找到

EdisionWew · 2021-08-17T01:56:20Z

我现在打算用train_custom.md里的那个mini-coco128试一试，但是那是在google - drive上的我又下不下来:-(，不过我看了一下这个mini-coco128的文件结构和我转化后的coco格式数据集是一样的，真的不知道是怎么回事，而且我昨天还用那个训练后的权重试了一下视频推理，平均每帧需要0.3s，我是用的yolox-s，我用yolov5s的话每帧平均7ms，这让我也很吃惊

这个我试过了，也是评估为0

Hezhexi2002 · 2021-08-17T02:20:38Z

我现在打算用train_custom.md里的那个mini-coco128试一试，但是那是在google - drive上的我又下不下来:-(，不过我看了一下这个mini-coco128的文件结构和我转化后的coco格式数据集是一样的，真的不知道是怎么回事，而且我昨天还用那个训练后的权重试了一下视频推理，平均每帧需要0.3s，我是用的yolox-s，我用yolov5s的话每帧平均7ms，这让我也很吃惊

这个我试过了，也是评估为0

啊这

Robert-Hopkins · 2021-08-17T02:35:56Z

请问解决了吗

MangoloD · 2021-08-17T06:35:11Z

我也遇到了，loss是正常下降的，但是ap就是0

所以你解决了吗？你也是coco格式数据集训练的吗？

我尝试过Coco数据集，是正常的，你可以检查下格式是否正确，同时我也像voc一样写了一篇文章

songtf525 · 2021-08-17T06:50:45Z

我也遇到了，loss是正常下降的，但是ap就是0

所以你解决了吗？你也是coco格式数据集训练的吗？

我尝试过Coco数据集，是正常的，你可以检查下格式是否正确，同时我也像voc一样写了一篇文章

多少迭代ap才有值？我迭代了10epoch，ap=0，我自己的数据集用的YOLO2COCO那个转的，我自己试了一下，没有问题，文章链接在哪？

MangoloD · 2021-08-17T07:25:06Z

我也遇到了，loss是正常下降的，但是ap就是0

所以你解决了吗？你也是coco格式数据集训练的吗？

我尝试过Coco数据集，是正常的，你可以检查下格式是否正确，同时我也像voc一样写了一篇文章

多少迭代ap才有值？我迭代了10epoch，ap=0，我自己的数据集用的YOLO2COCO那个转的，我自己试了一下，没有问题，文章链接在哪？

YOLOX训练COCO数据集，但我觉问题可能还是出在了转换上

MangoloD · 2021-08-17T07:26:59Z

I have met same problem, but my datasets are in VOC format.

2021-08-16 19:13:27 | INFO     | yolox.evaluators.voc_evaluator:160 - Evaluate in main process...
Writing stain VOC results file
Writing flaw VOC results file
Writing burn VOC results file
Eval IoU : 0.50
AP for stain = 0.0000
AP for flaw = 0.0000
AP for burn = 0.0000
Mean AP = 0.0000
~~~~~~~~
Results:
0.000
0.000
0.000
0.000
~~~~~~~~

2021-08-16 19:13:27 | INFO     | yolox.core.trainer:321 - 
Average forward time: 0.00 ms, Average NMS time: 0.00 ms, Average inference time: 0.00 ms

顺便问一下你解决了这个问题吗？我的数据在YOLO5上是可以训练的

请查看解决YOLOX训练时AP为0

Hezhexi2002 · 2021-08-17T14:56:28Z

I have met same problem, but my datasets are in VOC format.

2021-08-16 19:13:27 | INFO     | yolox.evaluators.voc_evaluator:160 - Evaluate in main process...
Writing stain VOC results file
Writing flaw VOC results file
Writing burn VOC results file
Eval IoU : 0.50
AP for stain = 0.0000
AP for flaw = 0.0000
AP for burn = 0.0000
Mean AP = 0.0000
~~~~~~~~
Results:
0.000
0.000
0.000
0.000
~~~~~~~~

2021-08-16 19:13:27 | INFO     | yolox.core.trainer:321 - 
Average forward time: 0.00 ms, Average NMS time: 0.00 ms, Average inference time: 0.00 ms

顺便问一下你解决了这个问题吗？我的数据在YOLO5上是可以训练的

请查看解决YOLOX训练时AP为0

所以是需要用xml转json吗？那这样我就得先yolo2voc再voc2coco吗？

MangoloD · 2021-08-18T02:16:49Z

I have met same problem, but my datasets are in VOC format.
2021-08-16 19:13:27 | INFO     | yolox.evaluators.voc_evaluator:160 - Evaluate in main process...
Writing stain VOC results file
Writing flaw VOC results file
Writing burn VOC results file
Eval IoU : 0.50
AP for stain = 0.0000
AP for flaw = 0.0000
AP for burn = 0.0000
Mean AP = 0.0000
~~~~~~~~
Results:
0.000
0.000
0.000
0.000
~~~~~~~~

2021-08-16 19:13:27 | INFO     | yolox.core.trainer:321 - 
Average forward time: 0.00 ms, Average NMS time: 0.00 ms, Average inference time: 0.00 ms
顺便问一下你解决了这个问题吗？我的数据在YOLO5上是可以训练的
请查看解决YOLOX训练时AP为0
所以是需要用xml转json吗？那这样我就得先yolo2voc再voc2coco吗？

不用吧，你只要确保转换出来的Coco数据格式是正确的就行

MangoloD · 2021-08-18T02:17:49Z

I have met same problem, but my datasets are in VOC format.
2021-08-16 19:13:27 | INFO     | yolox.evaluators.voc_evaluator:160 - Evaluate in main process...
Writing stain VOC results file
Writing flaw VOC results file
Writing burn VOC results file
Eval IoU : 0.50
AP for stain = 0.0000
AP for flaw = 0.0000
AP for burn = 0.0000
Mean AP = 0.0000
~~~~~~~~
Results:
0.000
0.000
0.000
0.000
~~~~~~~~

2021-08-16 19:13:27 | INFO     | yolox.core.trainer:321 - 
Average forward time: 0.00 ms, Average NMS time: 0.00 ms, Average inference time: 0.00 ms
顺便问一下你解决了这个问题吗？我的数据在YOLO5上是可以训练的
请查看解决YOLOX训练时AP为0
所以是需要用xml转json吗？那这样我就得先yolo2voc再voc2coco吗？

还有就是我这边也已经试验过Yolo数据格式的训练了，也可以正常跑通了

Hezhexi2002 · 2021-08-18T02:23:23Z

I have met same problem, but my datasets are in VOC format.
2021-08-16 19:13:27 | INFO     | yolox.evaluators.voc_evaluator:160 - Evaluate in main process...
Writing stain VOC results file
Writing flaw VOC results file
Writing burn VOC results file
Eval IoU : 0.50
AP for stain = 0.0000
AP for flaw = 0.0000
AP for burn = 0.0000
Mean AP = 0.0000
~~~~~~~~
Results:
0.000
0.000
0.000
0.000
~~~~~~~~

2021-08-16 19:13:27 | INFO     | yolox.core.trainer:321 - 
Average forward time: 0.00 ms, Average NMS time: 0.00 ms, Average inference time: 0.00 ms
顺便问一下你解决了这个问题吗？我的数据在YOLO5上是可以训练的
请查看解决YOLOX训练时AP为0
所以是需要用xml转json吗？那这样我就得先yolo2voc再voc2coco吗？
还有就是我这边也已经试验过Yolo数据格式的训练了，也可以正常跑通了

Hezhexi2002 · 2021-08-18T02:23:56Z

I have met same problem, but my datasets are in VOC format.
2021-08-16 19:13:27 | INFO     | yolox.evaluators.voc_evaluator:160 - Evaluate in main process...
Writing stain VOC results file
Writing flaw VOC results file
Writing burn VOC results file
Eval IoU : 0.50
AP for stain = 0.0000
AP for flaw = 0.0000
AP for burn = 0.0000
Mean AP = 0.0000
~~~~~~~~
Results:
0.000
0.000
0.000
0.000
~~~~~~~~

2021-08-16 19:13:27 | INFO     | yolox.core.trainer:321 - 
Average forward time: 0.00 ms, Average NMS time: 0.00 ms, Average inference time: 0.00 ms
顺便问一下你解决了这个问题吗？我的数据在YOLO5上是可以训练的
请查看解决YOLOX训练时AP为0
所以是需要用xml转json吗？那这样我就得先yolo2voc再voc2coco吗？
还有就是我这边也已经试验过Yolo数据格式的训练了，也可以正常跑通了

意思是直接用yolo2coco也行了吗？

songtf525 · 2021-08-18T02:50:42Z

我用同样的数据，yolo2voc后，ap不在等于0，但是调用作者给出链接的YOLO2COCO，仍旧不能用，所以，是否YOLO2COCO给出的json有问题？我对比了COCO的json，在key上给出的没有问题，有人测试过有不同的结果么？

Hezhexi2002 · 2021-08-18T02:52:18Z

我用同样的数据，yolo2voc后，ap不在等于0，但是调用作者给出链接的YOLO2COCO，仍旧不能用，所以，是否YOLO2COCO给出的json有问题？我对比了COCO的json，在key上给出的没有问题，有人测试过有不同的结果么？

我也是调用的作者给的yolo2coco然后ap一直为0

Hezhexi2002 · 2021-08-18T02:52:56Z

我用同样的数据，yolo2voc后，ap不在等于0，但是调用作者给出链接的YOLO2COCO，仍旧不能用，所以，是否YOLO2COCO给出的json有问题？我对比了COCO的json，在key上给出的没有问题，有人测试过有不同的结果么？

我也是调用的作者给的yolo2coco然后ap一直为0，我看B站上有人说了这是他们代码的问题，导致读取coco格式数据集时是根据id去拼的而不是依据imagepath

Robert-Hopkins · 2021-08-18T02:57:17Z

我用同样的数据，yolo2voc后，ap不在等于0，但是调用作者给出链接的YOLO2COCO，仍旧不能用，所以，是否YOLO2COCO给出的json有问题？我对比了COCO的json，在key上给出的没有问题，有人测试过有不同的结果么？

我也是调用的作者给的yolo2coco然后ap一直为0

YOLO2COCO代码有个地方有问题，所以导致AP=0，ann里面的id不对，需要修改

Hezhexi2002 · 2021-08-18T03:00:00Z

我用同样的数据，yolo2voc后，ap不在等于0，但是调用作者给出链接的YOLO2COCO，仍旧不能用，所以，是否YOLO2COCO给出的json有问题？我对比了COCO的json，在key上给出的没有问题，有人测试过有不同的结果么？

我也是调用的作者给的yolo2coco然后ap一直为0

YOLO2COCO代码有个地方有问题，所以导致AP=0，ann里面的id不对，需要修改

`# !/usr/bin/env python

-- encoding: utf-8 --

@file: yolov5_2_coco.py

@author: SWHL

@Contact: liekkaskono@163.com

import argparse
from pathlib import Path
import json
import shutil

import cv2 as cv

def read_txt(txt_path):
with open(str(txt_path), 'r', encoding='utf-8') as f:
data = f.readlines()
data = list(map(lambda x: x.rstrip('\n'), data))
return data

def mkdir(dir_path):
Path(dir_path).mkdir(parents=True, exist_ok=True)

class YOLOV5ToCOCO(object):
def init(self, dir_path):
self.src_data = Path(dir_path)
self.src = self.src_data.parent
self.train_txt_path = self.src_data / 'train.txt'
self.val_txt_path = self.src_data / 'val.txt'

    # 构建COCO格式目录
    self.dst = Path(self.src) / f"{Path(self.src_data).name}_COCO_format"
    self.coco_train = "train2017"
    self.coco_val = "val2017"
    self.coco_annotation = "annotations"
    self.coco_train_json = self.dst / self.coco_annotation \
                               / f'instances_{self.coco_train}.json'
    self.coco_val_json = self.dst / self.coco_annotation \
                               / f'instances_{self.coco_val}.json'

    mkdir(self.dst)
    mkdir(self.dst / self.coco_train)
    mkdir(self.dst / self.coco_val)
    mkdir(self.dst / self.coco_annotation)

    # 构建json内容结构
    self.type = 'instances'
    self.categories = []

    # 读取类别数
    self._get_category()

    self.info = {
        'year': 2021,
        'version': '1.0',
        'description': 'For object detection',
        'date_created': '2021',
    }

    self.licenses = [{
        'id': 1,
        'name': 'GNU General Public License v3.0',
        'url': 'https://github.com/zhiqwang/yolov5-rt-stack/blob/master/LICENSE',
    }]

def _get_category(self):
    class_list = read_txt(self.src_data / 'classes.txt')
    for i, category in enumerate(class_list, 1):
        self.categories.append({
            'id': i,
            'name': category,
            'supercategory': category,
        })

def generate(self):
    self.train_files = read_txt(self.train_txt_path)
    if Path(self.val_txt_path).exists():
        self.valid_files = read_txt(self.val_txt_path)

    train_dest_dir = Path(self.dst) / self.coco_train
    self.gen_dataset(self.train_files, train_dest_dir,
                     self.coco_train_json)

    val_dest_dir = Path(self.dst) / self.coco_val
    if Path(self.val_txt_path).exists():
        self.gen_dataset(self.valid_files, val_dest_dir,
                         self.coco_val_json)

    print(f"The output directory is: {str(self.dst)}")

def gen_dataset(self, img_paths, target_img_path, target_json):
    """
    https://cocodataset.org/#format-data

    """
    images = []
    annotations = []
    annotation_id = 1
    for img_id, img_path in enumerate(img_paths, 1):
        img_path = Path(img_path)

        if not img_path.exists():
            continue

        label_path = str(img_path.parent.parent
                         / 'labels' / f'{img_path.stem}.txt')

        imgsrc = cv.imread(str(img_path))
        height, width = imgsrc.shape[:2]

        dest_file_name = f'{img_id:012d}.jpg'
        save_img_path = target_img_path / dest_file_name

        if img_path.suffix.lower() == ".jpg":
            shutil.copyfile(img_path, save_img_path)
        else:
            cv.imwrite(str(save_img_path), imgsrc)

        images.append({
            'date_captured': '2021',
            'file_name': dest_file_name,
            'id': img_id,
            'height': height,
            'width': width,
        })

        if Path(label_path).exists():
            new_anno = self.read_annotation(label_path, img_id,
                                            height, width,
                                            annotation_id)
            if len(new_anno) > 0:
                annotations.extend(new_anno)
            else:
                raise ValueError(f'{label_path} is empty')
        else:
            raise FileExistsError(f'{label_path} not exists')

    json_data = {
        'info': self.info,
        'images': images,
        'licenses': self.licenses,
        'type': self.type,
        'annotations': annotations,
        'categories': self.categories,
    }
    with open(target_json, 'w', encoding='utf-8') as f:
        json.dump(json_data, f, ensure_ascii=False)

def read_annotation(self, txtfile, img_id,
                    height, width, annotation_id):
    annotation = []
    allinfo = read_txt(txtfile)
    for label_info in allinfo:
        label_info = label_info.split(" ")
        if len(label_info) < 5:
            continue

        category_id, vertex_info = label_info[0], label_info[1:]
        segmentation, bbox, area = self._get_annotation(vertex_info,
                                                        height, width)
        annotation.append({
            'segmentation': segmentation,
            'area': area,
            'iscrowd': 0,
            'image_id': img_id,
            'bbox': bbox,
            'category_id': int(category_id)+1,
            'id': annotation_id,
        })
        annotation_id += 1
    return annotation

@staticmethod
def _get_annotation(vertex_info, height, width):
    cx, cy, w, h = [float(i) for i in vertex_info]

    cx = cx * width
    cy = cy * height
    box_w = w * width
    box_h = h * height

    # left top
    x0 = max(cx - box_w / 2, 0)
    y0 = max(cy - box_h / 2, 0)

    # right bottomt
    x1 = min(x0 + box_w, width)
    y1 = min(y0 + box_h, height)

    segmentation = [[x0, y0, x1, y0, x1, y1, x0, y1]]
    bbox = [x0, y0, box_w, box_h]
    area = box_w * box_h
    return segmentation, bbox, area

if name == "main":
parser = argparse.ArgumentParser('Datasets converter from YOLOV5 to COCO')
parser.add_argument('--dir_path', type=str,
default='datasets/tmp/YOLOV5',
help='Dataset root path')
args = parser.parse_args()

converter = YOLOV5ToCOCO(args.dir_path)
converter.generate()

`
是哪里有问题？

是不是id应该从0开始？

Robert-Hopkins · 2021-08-18T03:01:42Z

我用同样的数据，yolo2voc后，ap不在等于0，但是调用作者给出链接的YOLO2COCO，仍旧不能用，所以，是否YOLO2COCO给出的json有问题？我对比了COCO的json，在key上给出的没有问题，有人测试过有不同的结果么？

我也是调用的作者给的yolo2coco然后ap一直为0

YOLO2COCO代码有个地方有问题，所以导致AP=0，ann里面的id不对，需要修改

我用同样的数据，yolo2voc后，ap不在等于0，但是调用作者给出链接的YOLO2COCO，仍旧不能用，所以，是否YOLO2COCO给出的json有问题？我对比了COCO的json，在key上给出的没有问题，有人测试过有不同的结果么？

我也是调用的作者给的yolo2coco然后ap一直为0

YOLO2COCO代码有个地方有问题，所以导致AP=0，ann里面的id不对，需要修改

`# !/usr/bin/env python

-- encoding: utf-8 --

@file: yolov5_2_coco.py

@author: SWHL

@Contact: liekkaskono@163.com

import argparse
from pathlib import Path
import json
import shutil

import cv2 as cv

def read_txt(txt_path):
with open(str(txt_path), 'r', encoding='utf-8') as f:
data = f.readlines()
data = list(map(lambda x: x.rstrip('\n'), data))
return data

def mkdir(dir_path):
Path(dir_path).mkdir(parents=True, exist_ok=True)

class YOLOV5ToCOCO(object):
def init(self, dir_path):
self.src_data = Path(dir_path)
self.src = self.src_data.parent
self.train_txt_path = self.src_data / 'train.txt'
self.val_txt_path = self.src_data / 'val.txt'

    # 构建COCO格式目录
    self.dst = Path(self.src) / f"{Path(self.src_data).name}_COCO_format"
    self.coco_train = "train2017"
    self.coco_val = "val2017"
    self.coco_annotation = "annotations"
    self.coco_train_json = self.dst / self.coco_annotation \
                               / f'instances_{self.coco_train}.json'
    self.coco_val_json = self.dst / self.coco_annotation \
                               / f'instances_{self.coco_val}.json'

    mkdir(self.dst)
    mkdir(self.dst / self.coco_train)
    mkdir(self.dst / self.coco_val)
    mkdir(self.dst / self.coco_annotation)

    # 构建json内容结构
    self.type = 'instances'
    self.categories = []

    # 读取类别数
    self._get_category()

    self.info = {
        'year': 2021,
        'version': '1.0',
        'description': 'For object detection',
        'date_created': '2021',
    }

    self.licenses = [{
        'id': 1,
        'name': 'GNU General Public License v3.0',
        'url': 'https://github.com/zhiqwang/yolov5-rt-stack/blob/master/LICENSE',
    }]

def _get_category(self):
    class_list = read_txt(self.src_data / 'classes.txt')
    for i, category in enumerate(class_list, 1):
        self.categories.append({
            'id': i,
            'name': category,
            'supercategory': category,
        })

def generate(self):
    self.train_files = read_txt(self.train_txt_path)
    if Path(self.val_txt_path).exists():
        self.valid_files = read_txt(self.val_txt_path)

    train_dest_dir = Path(self.dst) / self.coco_train
    self.gen_dataset(self.train_files, train_dest_dir,
                     self.coco_train_json)

    val_dest_dir = Path(self.dst) / self.coco_val
    if Path(self.val_txt_path).exists():
        self.gen_dataset(self.valid_files, val_dest_dir,
                         self.coco_val_json)

    print(f"The output directory is: {str(self.dst)}")

def gen_dataset(self, img_paths, target_img_path, target_json):
    """
    https://cocodataset.org/#format-data

    """
    images = []
    annotations = []
    annotation_id = 1
    for img_id, img_path in enumerate(img_paths, 1):
        img_path = Path(img_path)

        if not img_path.exists():
            continue

        label_path = str(img_path.parent.parent
                         / 'labels' / f'{img_path.stem}.txt')

        imgsrc = cv.imread(str(img_path))
        height, width = imgsrc.shape[:2]

        dest_file_name = f'{img_id:012d}.jpg'
        save_img_path = target_img_path / dest_file_name

        if img_path.suffix.lower() == ".jpg":
            shutil.copyfile(img_path, save_img_path)
        else:
            cv.imwrite(str(save_img_path), imgsrc)

        images.append({
            'date_captured': '2021',
            'file_name': dest_file_name,
            'id': img_id,
            'height': height,
            'width': width,
        })

        if Path(label_path).exists():
            new_anno = self.read_annotation(label_path, img_id,
                                            height, width,
                                            annotation_id)
            if len(new_anno) > 0:
                annotations.extend(new_anno)
            else:
                raise ValueError(f'{label_path} is empty')
        else:
            raise FileExistsError(f'{label_path} not exists')

    json_data = {
        'info': self.info,
        'images': images,
        'licenses': self.licenses,
        'type': self.type,
        'annotations': annotations,
        'categories': self.categories,
    }
    with open(target_json, 'w', encoding='utf-8') as f:
        json.dump(json_data, f, ensure_ascii=False)

def read_annotation(self, txtfile, img_id,
                    height, width, annotation_id):
    annotation = []
    allinfo = read_txt(txtfile)
    for label_info in allinfo:
        label_info = label_info.split(" ")
        if len(label_info) < 5:
            continue

        category_id, vertex_info = label_info[0], label_info[1:]
        segmentation, bbox, area = self._get_annotation(vertex_info,
                                                        height, width)
        annotation.append({
            'segmentation': segmentation,
            'area': area,
            'iscrowd': 0,
            'image_id': img_id,
            'bbox': bbox,
            'category_id': int(category_id)+1,
            'id': annotation_id,
        })
        annotation_id += 1
    return annotation

@staticmethod
def _get_annotation(vertex_info, height, width):
    cx, cy, w, h = [float(i) for i in vertex_info]

    cx = cx * width
    cy = cy * height
    box_w = w * width
    box_h = h * height

    # left top
    x0 = max(cx - box_w / 2, 0)
    y0 = max(cy - box_h / 2, 0)

    # right bottomt
    x1 = min(x0 + box_w, width)
    y1 = min(y0 + box_h, height)

    segmentation = [[x0, y0, x1, y0, x1, y1, x0, y1]]
    bbox = [x0, y0, box_w, box_h]
    area = box_w * box_h
    return segmentation, bbox, area

if name == "main":
parser = argparse.ArgumentParser('Datasets converter from YOLOV5 to COCO')
parser.add_argument('--dir_path', type=str,
default='datasets/tmp/YOLOV5',
help='Dataset root path')
args = parser.parse_args()

converter = YOLOV5ToCOCO(args.dir_path)
converter.generate()

`
是哪里有问题？YOLO2COCO代码有个地方有问题，所以导致AP=0，ann里面的id不对，需要修改

Robert-Hopkins · 2021-08-18T03:02:07Z

YOLO2COCO代码有个地方有问题，所以导致AP=0，ann里面的id不对，需要修改

Robert-Hopkins · 2021-08-18T03:26:33Z

我用同样的数据，yolo2voc后，ap不在等于0，但是调用作者给出链接的YOLO2COCO，仍旧不能用，所以，是否YOLO2COCO给出的json有问题？我对比了COCO的json，在key上给出的没有问题，有人测试过有不同的结果么？

我也是调用的作者给的yolo2coco然后ap一直为0

YOLO2COCO代码有个地方有问题，所以导致AP=0，ann里面的id不对，需要修改

`# !/usr/bin/env python

-- encoding: utf-8 --

@file: yolov5_2_coco.py

@author: SWHL

@Contact: liekkaskono@163.com

import argparse
from pathlib import Path
import json
import shutil

import cv2 as cv

def read_txt(txt_path):
with open(str(txt_path), 'r', encoding='utf-8') as f:
data = f.readlines()
data = list(map(lambda x: x.rstrip('\n'), data))
return data

def mkdir(dir_path):
Path(dir_path).mkdir(parents=True, exist_ok=True)

class YOLOV5ToCOCO(object):
def init(self, dir_path):
self.src_data = Path(dir_path)
self.src = self.src_data.parent
self.train_txt_path = self.src_data / 'train.txt'
self.val_txt_path = self.src_data / 'val.txt'

    # 构建COCO格式目录
    self.dst = Path(self.src) / f"{Path(self.src_data).name}_COCO_format"
    self.coco_train = "train2017"
    self.coco_val = "val2017"
    self.coco_annotation = "annotations"
    self.coco_train_json = self.dst / self.coco_annotation \
                               / f'instances_{self.coco_train}.json'
    self.coco_val_json = self.dst / self.coco_annotation \
                               / f'instances_{self.coco_val}.json'

    mkdir(self.dst)
    mkdir(self.dst / self.coco_train)
    mkdir(self.dst / self.coco_val)
    mkdir(self.dst / self.coco_annotation)

    # 构建json内容结构
    self.type = 'instances'
    self.categories = []

    # 读取类别数
    self._get_category()

    self.info = {
        'year': 2021,
        'version': '1.0',
        'description': 'For object detection',
        'date_created': '2021',
    }

    self.licenses = [{
        'id': 1,
        'name': 'GNU General Public License v3.0',
        'url': 'https://github.com/zhiqwang/yolov5-rt-stack/blob/master/LICENSE',
    }]

def _get_category(self):
    class_list = read_txt(self.src_data / 'classes.txt')
    for i, category in enumerate(class_list, 1):
        self.categories.append({
            'id': i,
            'name': category,
            'supercategory': category,
        })

def generate(self):
    self.train_files = read_txt(self.train_txt_path)
    if Path(self.val_txt_path).exists():
        self.valid_files = read_txt(self.val_txt_path)

    train_dest_dir = Path(self.dst) / self.coco_train
    self.gen_dataset(self.train_files, train_dest_dir,
                     self.coco_train_json)

    val_dest_dir = Path(self.dst) / self.coco_val
    if Path(self.val_txt_path).exists():
        self.gen_dataset(self.valid_files, val_dest_dir,
                         self.coco_val_json)

    print(f"The output directory is: {str(self.dst)}")

def gen_dataset(self, img_paths, target_img_path, target_json):
    """
    https://cocodataset.org/#format-data

    """
    images = []
    annotations = []
    annotation_id = 1
    for img_id, img_path in enumerate(img_paths, 1):
        img_path = Path(img_path)

        if not img_path.exists():
            continue

        label_path = str(img_path.parent.parent
                         / 'labels' / f'{img_path.stem}.txt')

        imgsrc = cv.imread(str(img_path))
        height, width = imgsrc.shape[:2]

        dest_file_name = f'{img_id:012d}.jpg'
        save_img_path = target_img_path / dest_file_name

        if img_path.suffix.lower() == ".jpg":
            shutil.copyfile(img_path, save_img_path)
        else:
            cv.imwrite(str(save_img_path), imgsrc)

        images.append({
            'date_captured': '2021',
            'file_name': dest_file_name,
            'id': img_id,
            'height': height,
            'width': width,
        })

        if Path(label_path).exists():
            new_anno = self.read_annotation(label_path, img_id,
                                            height, width,
                                            annotation_id)
            if len(new_anno) > 0:
                annotations.extend(new_anno)
            else:
                raise ValueError(f'{label_path} is empty')
        else:
            raise FileExistsError(f'{label_path} not exists')

    json_data = {
        'info': self.info,
        'images': images,
        'licenses': self.licenses,
        'type': self.type,
        'annotations': annotations,
        'categories': self.categories,
    }
    with open(target_json, 'w', encoding='utf-8') as f:
        json.dump(json_data, f, ensure_ascii=False)

def read_annotation(self, txtfile, img_id,
                    height, width, annotation_id):
    annotation = []
    allinfo = read_txt(txtfile)
    for label_info in allinfo:
        label_info = label_info.split(" ")
        if len(label_info) < 5:
            continue

        category_id, vertex_info = label_info[0], label_info[1:]
        segmentation, bbox, area = self._get_annotation(vertex_info,
                                                        height, width)
        annotation.append({
            'segmentation': segmentation,
            'area': area,
            'iscrowd': 0,
            'image_id': img_id,
            'bbox': bbox,
            'category_id': int(category_id)+1,
            'id': annotation_id,
        })
        annotation_id += 1
    return annotation

@staticmethod
def _get_annotation(vertex_info, height, width):
    cx, cy, w, h = [float(i) for i in vertex_info]

    cx = cx * width
    cy = cy * height
    box_w = w * width
    box_h = h * height

    # left top
    x0 = max(cx - box_w / 2, 0)
    y0 = max(cy - box_h / 2, 0)

    # right bottomt
    x1 = min(x0 + box_w, width)
    y1 = min(y0 + box_h, height)

    segmentation = [[x0, y0, x1, y0, x1, y1, x0, y1]]
    bbox = [x0, y0, box_w, box_h]
    area = box_w * box_h
    return segmentation, bbox, area

if name == "main":
parser = argparse.ArgumentParser('Datasets converter from YOLOV5 to COCO')
parser.add_argument('--dir_path', type=str,
default='datasets/tmp/YOLOV5',
help='Dataset root path')
args = parser.parse_args()

converter = YOLOV5ToCOCO(args.dir_path)
converter.generate()

`
是哪里有问题？

是不是id应该从0开始？

不是，他的id不唯一，每个图片的id都是从1开始的，所以有重复

Hezhexi2002 · 2021-08-18T03:32:00Z

我用同样的数据，yolo2voc后，ap不在等于0，但是调用作者给出链接的YOLO2COCO，仍旧不能用，所以，是否YOLO2COCO给出的json有问题？我对比了COCO的json，在key上给出的没有问题，有人测试过有不同的结果么？

我也是调用的作者给的yolo2coco然后ap一直为0

YOLO2COCO代码有个地方有问题，所以导致AP=0，ann里面的id不对，需要修改

`# !/usr/bin/env python

-- encoding: utf-8 --

@file: yolov5_2_coco.py

@author: SWHL

@Contact: liekkaskono@163.com

import argparse
from pathlib import Path
import json
import shutil

import cv2 as cv

def read_txt(txt_path):
with open(str(txt_path), 'r', encoding='utf-8') as f:
data = f.readlines()
data = list(map(lambda x: x.rstrip('\n'), data))
return data

def mkdir(dir_path):
Path(dir_path).mkdir(parents=True, exist_ok=True)

class YOLOV5ToCOCO(object):
def init(self, dir_path):
self.src_data = Path(dir_path)
self.src = self.src_data.parent
self.train_txt_path = self.src_data / 'train.txt'
self.val_txt_path = self.src_data / 'val.txt'

    # 构建COCO格式目录
    self.dst = Path(self.src) / f"{Path(self.src_data).name}_COCO_format"
    self.coco_train = "train2017"
    self.coco_val = "val2017"
    self.coco_annotation = "annotations"
    self.coco_train_json = self.dst / self.coco_annotation \
                               / f'instances_{self.coco_train}.json'
    self.coco_val_json = self.dst / self.coco_annotation \
                               / f'instances_{self.coco_val}.json'

    mkdir(self.dst)
    mkdir(self.dst / self.coco_train)
    mkdir(self.dst / self.coco_val)
    mkdir(self.dst / self.coco_annotation)

    # 构建json内容结构
    self.type = 'instances'
    self.categories = []

    # 读取类别数
    self._get_category()

    self.info = {
        'year': 2021,
        'version': '1.0',
        'description': 'For object detection',
        'date_created': '2021',
    }

    self.licenses = [{
        'id': 1,
        'name': 'GNU General Public License v3.0',
        'url': 'https://github.com/zhiqwang/yolov5-rt-stack/blob/master/LICENSE',
    }]

def _get_category(self):
    class_list = read_txt(self.src_data / 'classes.txt')
    for i, category in enumerate(class_list, 1):
        self.categories.append({
            'id': i,
            'name': category,
            'supercategory': category,
        })

def generate(self):
    self.train_files = read_txt(self.train_txt_path)
    if Path(self.val_txt_path).exists():
        self.valid_files = read_txt(self.val_txt_path)

    train_dest_dir = Path(self.dst) / self.coco_train
    self.gen_dataset(self.train_files, train_dest_dir,
                     self.coco_train_json)

    val_dest_dir = Path(self.dst) / self.coco_val
    if Path(self.val_txt_path).exists():
        self.gen_dataset(self.valid_files, val_dest_dir,
                         self.coco_val_json)

    print(f"The output directory is: {str(self.dst)}")

def gen_dataset(self, img_paths, target_img_path, target_json):
    """
    https://cocodataset.org/#format-data

    """
    images = []
    annotations = []
    annotation_id = 1
    for img_id, img_path in enumerate(img_paths, 1):
        img_path = Path(img_path)

        if not img_path.exists():
            continue

        label_path = str(img_path.parent.parent
                         / 'labels' / f'{img_path.stem}.txt')

        imgsrc = cv.imread(str(img_path))
        height, width = imgsrc.shape[:2]

        dest_file_name = f'{img_id:012d}.jpg'
        save_img_path = target_img_path / dest_file_name

        if img_path.suffix.lower() == ".jpg":
            shutil.copyfile(img_path, save_img_path)
        else:
            cv.imwrite(str(save_img_path), imgsrc)

        images.append({
            'date_captured': '2021',
            'file_name': dest_file_name,
            'id': img_id,
            'height': height,
            'width': width,
        })

        if Path(label_path).exists():
            new_anno = self.read_annotation(label_path, img_id,
                                            height, width,
                                            annotation_id)
            if len(new_anno) > 0:
                annotations.extend(new_anno)
            else:
                raise ValueError(f'{label_path} is empty')
        else:
            raise FileExistsError(f'{label_path} not exists')

    json_data = {
        'info': self.info,
        'images': images,
        'licenses': self.licenses,
        'type': self.type,
        'annotations': annotations,
        'categories': self.categories,
    }
    with open(target_json, 'w', encoding='utf-8') as f:
        json.dump(json_data, f, ensure_ascii=False)

def read_annotation(self, txtfile, img_id,
                    height, width, annotation_id):
    annotation = []
    allinfo = read_txt(txtfile)
    for label_info in allinfo:
        label_info = label_info.split(" ")
        if len(label_info) < 5:
            continue

        category_id, vertex_info = label_info[0], label_info[1:]
        segmentation, bbox, area = self._get_annotation(vertex_info,
                                                        height, width)
        annotation.append({
            'segmentation': segmentation,
            'area': area,
            'iscrowd': 0,
            'image_id': img_id,
            'bbox': bbox,
            'category_id': int(category_id)+1,
            'id': annotation_id,
        })
        annotation_id += 1
    return annotation

@staticmethod
def _get_annotation(vertex_info, height, width):
    cx, cy, w, h = [float(i) for i in vertex_info]

    cx = cx * width
    cy = cy * height
    box_w = w * width
    box_h = h * height

    # left top
    x0 = max(cx - box_w / 2, 0)
    y0 = max(cy - box_h / 2, 0)

    # right bottomt
    x1 = min(x0 + box_w, width)
    y1 = min(y0 + box_h, height)

    segmentation = [[x0, y0, x1, y0, x1, y1, x0, y1]]
    bbox = [x0, y0, box_w, box_h]
    area = box_w * box_h
    return segmentation, bbox, area

if name == "main":
parser = argparse.ArgumentParser('Datasets converter from YOLOV5 to COCO')
parser.add_argument('--dir_path', type=str,
default='datasets/tmp/YOLOV5',
help='Dataset root path')
args = parser.parse_args()

converter = YOLOV5ToCOCO(args.dir_path)
converter.generate()

`
是哪里有问题？

是不是id应该从0开始？

不是，他的id不唯一，每个图片的id都是从1开始的，所以有重复

所以你是怎样修改的呢？

songtf525 · 2021-08-18T06:58:24Z

针对YOLO2COCO的，id如果唯一，那就把 annotation_id = 1提升成全局变量，然后函数def read_annotation(self, txtfile, img_id, height, width)，把annotation_id去掉，调用时候也要去掉, 我改完试了一个数据集，it is work， very good，thankyou!

nTnZone · 2021-08-18T10:14:23Z

I have met same problem, but my datasets are in VOC format.
2021-08-16 19:13:27 | INFO     | yolox.evaluators.voc_evaluator:160 - Evaluate in main process...
Writing stain VOC results file
Writing flaw VOC results file
Writing burn VOC results file
Eval IoU : 0.50
AP for stain = 0.0000
AP for flaw = 0.0000
AP for burn = 0.0000
Mean AP = 0.0000
~~~~~~~~
Results:
0.000
0.000
0.000
0.000
~~~~~~~~

2021-08-16 19:13:27 | INFO     | yolox.core.trainer:321 - 
Average forward time: 0.00 ms, Average NMS time: 0.00 ms, Average inference time: 0.00 ms
顺便问一下你解决了这个问题吗？我的数据在YOLO5上是可以训练的
还没有：-（，但是voc格式mAP为0已经有解决方案了，你在issue可以找到

我解决了，应该是路径没设置正确。相比yolo5,这个得配置很多地方，有其实修改自定义数据集的路径，删除多余的信息如self.year之类的

Hezhexi2002 · 2021-08-18T10:42:31Z

针对YOLO2COCO的，id如果唯一，那就把 annotation_id = 1提升成全局变量，然后函数def read_annotation(self, txtfile, img_id, height, width)，把annotation_id去掉，调用时候也要去掉, 我改完试了一个数据集，it is work， very good，thankyou!

Hezhexi2002 · 2021-08-18T10:44:50Z

针对YOLO2COCO的，id如果唯一，那就把 annotation_id = 1提升成全局变量，然后函数def read_annotation(self, txtfile, img_id, height, width)，把annotation_id去掉，调用时候也要去掉, 我改完试了一个数据集，it is work， very good，thankyou!

我也试了确实有效，太感谢了！！！😭

Hezhexi2002 closed this as completed Aug 15, 2021

Hezhexi2002 reopened this Aug 15, 2021

Hezhexi2002 closed this as completed Aug 18, 2021

Hezhexi2002 reopened this Aug 18, 2021

Hezhexi2002 closed this as completed Aug 18, 2021

Hezhexi2002 reopened this Aug 18, 2021

Hezhexi2002 closed this as completed Aug 18, 2021

Hezhexi2002 reopened this Aug 18, 2021

Hezhexi2002 closed this as completed Aug 19, 2021

Joker316701882 pinned this issue Aug 22, 2021

Joker316701882 unpinned this issue Oct 15, 2021

The mAP is always 0 when I train on my custom data in coco format? #504

The mAP is always 0 when I train on my custom data in coco format? #504

Comments

Hezhexi2002 commented Aug 15, 2021 • edited

-- coding:utf-8 --

Copyright (c) Megvii, Inc. and its affiliates.

-- coding:utf-8 --

Copyright (c) Megvii, Inc. and its affiliates.

Bachelorwangwei commented Aug 16, 2021

Hezhexi2002 commented Aug 16, 2021 • edited

EdisionWew commented Aug 16, 2021

Hezhexi2002 commented Aug 16, 2021

Hezhexi2002 commented Aug 17, 2021 • edited

EdisionWew commented Aug 17, 2021

Hezhexi2002 commented Aug 17, 2021

Robert-Hopkins commented Aug 17, 2021

MangoloD commented Aug 17, 2021

songtf525 commented Aug 17, 2021

MangoloD commented Aug 17, 2021

MangoloD commented Aug 17, 2021

Hezhexi2002 commented Aug 17, 2021 • edited

MangoloD commented Aug 18, 2021

MangoloD commented Aug 18, 2021

Hezhexi2002 commented Aug 18, 2021

Hezhexi2002 commented Aug 18, 2021

songtf525 commented Aug 18, 2021

Hezhexi2002 commented Aug 18, 2021

Hezhexi2002 commented Aug 18, 2021 • edited

Robert-Hopkins commented Aug 18, 2021

Hezhexi2002 commented Aug 18, 2021 • edited

-- encoding: utf-8 --

@file: yolov5_2_coco.py

@author: SWHL

@Contact: liekkaskono@163.com

Robert-Hopkins commented Aug 18, 2021

-- encoding: utf-8 --

@file: yolov5_2_coco.py

@author: SWHL

@Contact: liekkaskono@163.com

Robert-Hopkins commented Aug 18, 2021

Robert-Hopkins commented Aug 18, 2021

-- encoding: utf-8 --

@file: yolov5_2_coco.py

@author: SWHL

@Contact: liekkaskono@163.com

Hezhexi2002 commented Aug 18, 2021

-- encoding: utf-8 --

@file: yolov5_2_coco.py

@author: SWHL

@Contact: liekkaskono@163.com

songtf525 commented Aug 18, 2021

nTnZone commented Aug 18, 2021

Hezhexi2002 commented Aug 18, 2021

Hezhexi2002 commented Aug 18, 2021

Hezhexi2002 commented Aug 15, 2021 •

edited

Hezhexi2002 commented Aug 16, 2021 •

edited

Hezhexi2002 commented Aug 17, 2021 •

edited

Hezhexi2002 commented Aug 17, 2021 •

edited

Hezhexi2002 commented Aug 18, 2021 •

edited

Hezhexi2002 commented Aug 18, 2021 •

edited