MODEL_ZOO.md

Model Zoo and Baselines

Hardware

• 8 NVIDIA V100 GPUs

Software

• PyTorch version: 1.0.0a0+dd2c487
• CUDA 9.2
• CUDNN 7.1
• NCCL 2.2.13-1

End-to-end Faster and Mask R-CNN baselines

All the baselines were trained using the exact same experimental setup as in Detectron. We initialize the detection models with ImageNet weights from Caffe2, the same as used by Detectron.

The pre-trained models are available in the link in the model id.

backbone	type	lr sched	im / gpu	train mem(GB)	train time (s/iter)	total train time(hr)	inference time(s/im)	box AP	mask AP	model id
R-50-C4	Fast	1x	1	5.8	0.4036	20.2	0.17130	34.8	-	6358800
R-50-FPN	Fast	1x	2	4.4	0.3530	8.8	0.12580	36.8	-	6358793
R-101-FPN	Fast	1x	2	7.1	0.4591	11.5	0.143149	39.1	-	6358804
X-101-32x8d-FPN	Fast	1x	1	7.6	0.7007	35.0	0.209965	41.2	-	6358717
R-50-C4	Mask	1x	1	5.8	0.4520	22.6	0.17796 + 0.028	35.6	31.5	6358801
R-50-FPN	Mask	1x	2	5.2	0.4536	11.3	0.12966 + 0.034	37.8	34.2	6358792
R-101-FPN	Mask	1x	2	7.9	0.5665	14.2	0.15384 + 0.034	40.1	36.1	6358805
X-101-32x8d-FPN	Mask	1x	1	7.8	0.7562	37.8	0.21739 + 0.034	42.2	37.8	6358718

For person keypoint detection:

backbone	type	lr sched	im / gpu	train mem(GB)	train time (s/iter)	total train time(hr)	inference time(s/im)	box AP	keypoint AP	model id
R-50-FPN	Keypoint	1x	2	5.7	0.3771	9.4	0.10941	53.7	64.3	9981060

Light-weight Model baselines

We provided pre-trained models for selected FBNet models.

• All the models are trained from scratched with BN using the training schedule specified below.
• Evaluation is performed on a single NVIDIA V100 GPU with MODEL.RPN.POST_NMS_TOP_N_TEST set to 200.

The following inference time is reported:

• inference total batch=8: Total inference time including data loading, model inference and pre/post preprocessing using 8 images per batch.
• inference model batch=8: Model inference time only and using 8 images per batch.
• inference model batch=1: Model inference time only and using 1 image per batch.
• inferenee caffe2 batch=1: Model inference time for the model in Caffe2 format using 1 image per batch. The Caffe2 models fused the BN to Conv and purely run on C++/CUDA by using Caffe2 ops for rpn/detection post processing.

The pre-trained models are available in the link in the model id.

backbone	type	resolution	lr sched	im / gpu	train mem(GB)	train time (s/iter)	total train time (hr)	inference total batch=8 (s/im)	inference model batch=8 (s/im)	inference model batch=1 (s/im)	inference caffe2 batch=1 (s/im)	box AP	mask AP	model id
R-50-C4 (reference)	Fast	800	1x	1	5.8	0.4036	20.2	0.0875	0.0793	0.0831	0.0625	34.4	-	f35857197
fbnet_chamv1a	Fast	600	0.75x	12	13.6	0.5444	20.5	0.0315	0.0260	0.0376	0.0188	33.5	-	f100940543
fbnet_default	Fast	600	0.5x	16	11.1	0.4872	12.5	0.0316	0.0250	0.0297	0.0130	28.2	-	f101086388
R-50-C4 (reference)	Mask	800	1x	1	5.8	0.452	22.6	0.0918	0.0848	0.0844	-	35.2	31.0	f35858791
fbnet_xirb16d	Mask	600	0.5x	16	13.4	1.1732	29	0.0386	0.0319	0.0356	-	30.7	26.9	f101086394
fbnet_default	Mask	600	0.5x	16	13.0	0.9036	23.0	0.0327	0.0269	0.0385	-	29.0	26.1	f101086385

Comparison with Detectron and mmdetection

In the following section, we compare our implementation with Detectron and mmdetection. The same remarks from mmdetection about different hardware applies here.

Training speed

The numbers here are in seconds / iteration. The lower, the better.

type	Detectron (P100)	mmdetection (V100)	maskrcnn_benchmark (V100)
Faster R-CNN R-50 C4	0.566	-	0.4036
Faster R-CNN R-50 FPN	0.544	0.554	0.3530
Faster R-CNN R-101 FPN	0.647	-	0.4591
Faster R-CNN X-101-32x8d FPN	0.799	-	0.7007
Mask R-CNN R-50 C4	0.620	-	0.4520
Mask R-CNN R-50 FPN	0.889	0.690	0.4536
Mask R-CNN R-101 FPN	1.008	-	0.5665
Mask R-CNN X-101-32x8d FPN	0.961	-	0.7562

Training memory

The lower, the better

type	Detectron (P100)	mmdetection (V100)	maskrcnn_benchmark (V100)
Faster R-CNN R-50 C4	6.3	-	5.8
Faster R-CNN R-50 FPN	7.2	4.9	4.4
Faster R-CNN R-101 FPN	8.9	-	7.1
Faster R-CNN X-101-32x8d FPN	7.0	-	7.6
Mask R-CNN R-50 C4	6.6	-	5.8
Mask R-CNN R-50 FPN	8.6	5.9	5.2
Mask R-CNN R-101 FPN	10.2	-	7.9
Mask R-CNN X-101-32x8d FPN	7.7	-	7.8

Accuracy

The higher, the better

type	Detectron (P100)	mmdetection (V100)	maskrcnn_benchmark (V100)
Faster R-CNN R-50 C4	34.8	-	34.8
Faster R-CNN R-50 FPN	36.7	36.7	36.8
Faster R-CNN R-101 FPN	39.4	-	39.1
Faster R-CNN X-101-32x8d FPN	41.3	-	41.2
Mask R-CNN R-50 C4	35.8 & 31.4	-	35.6 & 31.5
Mask R-CNN R-50 FPN	37.7 & 33.9	37.5 & 34.4	37.8 & 34.2
Mask R-CNN R-101 FPN	40.0 & 35.9	-	40.1 & 36.1
Mask R-CNN X-101-32x8d FPN	42.1 & 37.3	-	42.2 & 37.8