/
r50_pascalpartbase_voc.yaml
48 lines (48 loc) · 1.39 KB
/
r50_pascalpartbase_voc.yaml
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
_BASE_: "../coco/r50_coco.yaml"
MODEL:
ROI_HEADS:
NUM_CLASSES: 77 # max of num_classes in all train datasets
ROI_BOX_HEAD:
MULT_OBJECT_SCORE: True
USE_SIGMOID_CE: True
USE_ZEROSHOT_CLS: True
USE_ZEROSHOT_CLS_GROUP: True
ZEROSHOT_WEIGHT_DIM: 1024
ZEROSHOT_WEIGHT_PATH_GROUP: [
'datasets/metadata/pascal_part_base_clip_RN50_a+cname.npy',
'datasets/metadata/voc_clip_RN50_a+cname.npy'
]
ZEROSHOT_WEIGHT_INFERENCE_PATH: 'datasets/metadata/pascal_part_clip_RN50_a+cname.npy'
IGNORE_ZERO_CATS_GROUP: [False, False]
USE_FED_LOSS_GROUP: [False, False]
DATASETS:
TRAIN: ("pascal_part_base_train", "voc_2007_train")
TEST: ("pascal_part_val",)
SOLVER:
IMS_PER_BATCH: 16
USE_CUSTOM_SOLVER: True
OPTIMIZER: "ADAMW"
BASE_LR: 0.00005
STEPS: (12000, 16000)
MAX_ITER: 20000
CHECKPOINT_PERIOD: 5000
TEST:
DETECTIONS_PER_IMAGE: 100
EVAL_PERIOD: 5000
INPUT:
CUSTOM_AUG: ResizeShortestEdge
MIN_SIZE_TRAIN_SAMPLING: range
MIN_SIZE_TRAIN: (640, 800)
DATALOADER:
SAMPLER_TRAIN: "MultiDatasetSampler"
DATASET_RATIO: [1, 1]
USE_DIFF_BS_SIZE: True
DATASET_BS: [2, 2]
USE_RFS: [False, False]
DATASET_MIN_SIZES: [[640, 800], [640, 800]]
DATASET_MAX_SIZES: [1333, 1333]
FILTER_EMPTY_ANNOTATIONS: False
MULTI_DATASET_GROUPING: True
DATASET_ANN: ['part', 'box']
NUM_WORKERS: 8
WITH_IMAGE_LABELS: True