MODEL ZOO

November 8, 2022 ยท View on GitHub

MSCOCO dataset

ModelBackboneDetectorInput SizeAPSpeedDownloadConfigTraining Log
Simple BaselineResNet50YOLOv3256x19270.62.94 iter/smodelcfglog
Fast PoseResNet50YOLOv3256x19272.03.54 iter/smodelcfglog
Fast Pose (DUC)ResNet50 - unshuffleYOLOv3256x19272.42.91 iter/smodelcfglog
HRNetHRNet-W32YOLOv3256x19272.52.13 iter/smodelcfglog
Fast Pose (DCN)ResNet50 - dcnYOLOv3256x19272.82.94 iter/smodelcfglog
Fast Pose (DUC)ResNet152YOLOv3256x19273.31.62 iter/smodelcfglog

Notes

  • All models are trained on keypoint train 2017 images which contains at least one human with keypoint annotations (64115 images).
  • The evaluation is done on COCO keypoint val 2017 (5000 images).
  • Flip test is used by default.
  • One TITAN XP is used for speed test, with batch_size=64 in each iteration.
  • Offline human detection results are used in speed test.
  • FastPose is our own network design. Paper coming soon!

Halpe dataset (26 keypoints)

ModelBackboneDetectorInput SizeAPSpeedDownloadConfig
Fast PoseResNet50YOLOv3256x192-13.12 iter/sGoogle Baiducfg

For example, you can run with:

python scripts/demo_inference.py --cfg configs/halpe_26/resnet/256x192_res50_lr1e-3_1x.yaml --checkpoint pretrained_models/halpe26_fast_res50_256x192.pth --indir examples/demo/ --save_img

Notes

  • This model is trained based on the first 26 keypoints of Halpe Full-body datatset (without face and hand keypoints).
  • The speed is tested on COCO val2017 on a single NVIDIA GeForce RTX 3090 gpu, with batch_size=64 in each iteration and offline yolov3 human detection results.
ModelBackboneDetectorInput SizeLoss TypeAPSpeedDownloadConfig#keypoints
Fast PoseResNet50YOLOv3256x192Symmetric Integral50.116.28 iter/sGoogle Baidu(code: d0wi)cfg136
Fast Pose (DCN)ResNet50 - dcnYOLOv3256x192Combined (10 hand weight)49.810.35 iter/sGoogle Baidu(code: app1)cfg136
Fast Pose (DCN)ResNet50 - dcnYOLOv3256x192Combined-13.88 iter/sGoogle Baidu(code: 6kwr)cfg68 (no face)
Fast Pose (DCN)ResNet50 - dcn-256x192Symmetric Integral-30.20 iter/sGoogle Baidu(code: nwxx)cfg21 (single hand)

For the most accurate wholebody pose estimation, you can run with:

python scripts/demo_inference.py --cfg configs/halpe_coco_wholebody_136/resnet/256x192_res50_lr1e-3_2x-dcn-combined.yaml --checkpoint pretrained_models/multi_domain_fast50_dcn_combined_256x192.pth --indir examples/demo/ --save_img

or, you can run with (this version is a little faster and more accurate on body keypoints, but its performance on hand keypoints is worser):

python scripts/demo_inference.py --cfg configs/halpe_coco_wholebody_136/resnet/256x192_res50_lr1e-3_2x-regression.yaml --checkpoint pretrained_models/multi_domain_fast50_regression_256x192.pth --indir examples/demo/ --save_img

Notes

  • Above models are trained on multiple datasets, thus they can perform well for in-the-wild images.
  • 'Combined (10 hand weight)' means that we use different loss for hand and body keypoints

Halpe dataset (136 keypoints)

ModelBackboneDetectorInput SizeLoss TypeAPSpeedDownloadConfig
Fast PoseResNet50YOLOv3256x192Heatmap41.74.37 iter/sGoogle Baidu(code: y8a0)cfg
Fast PoseResNet50YOLOv3256x192Symmetric Integral44.116.50 iter/sGoogle Baidu(code: 9e4z)cfg
Fast Pose (DCN)ResNet50 - dcnYOLOv3256x192Symmetric Integral46.216.58 iter/sGoogle Baidu(code: 0yyf)cfg
Fast Pose (DCN)ResNet50 - dcnYOLOv3256x192Combined45.410.07 iter/sGoogle Baidu(code: hln3)cfg
Fast Pose (DCN)ResNet50 - dcnYOLOv3256x192Combined (10 hand weight)47.210.07 iter/sGoogle Baidu(code: jkyc)cfg
Fast Pose (DUC)ResNet152YOLOv3256x192Symmetric Integral45.116.17 iter/sGoogle Baidu(code: gaxj)cfg

For example, you can run with:

python scripts/demo_inference.py --cfg configs/halpe_136/resnet/256x192_res50_lr1e-3_2x-regression.yaml --checkpoint pretrained_models/halpe136_fast50_regression_256x192.pth --indir examples/demo/ --save_img

Notes

  • All of above models are trained only on Halpe Full-body dataset.
  • The APs are tested under Halpe's criterion, with flip test on.
  • Combined loss means we use heatmap loss (mse loss) on body and foot keypoints and use symmetric integral loss (l1 joint regression loss) on face and hand keypoints.
  • There are two FastPose-DCN models with combined loss. The second one uses ten times of weight of hand keypoints, so it is more accurate on hand keypoints but less accurate on the other keypoints.
  • The speed is tested on COCO val2017 on a single NVIDIA GeForce RTX 3090 gpu, with batch_size=64 in each iteration and offline yolov3 human detection results.

COCO WholeBody dataset (133 keypoints)

ModelBackboneDetectorInput SizeLoss TypeAPSpeedDownloadConfig
Fast PoseResNet50YOLOv3256x192Symmetric Integral55.417.42 iter/sGoogle Baidu(code: nw03)cfg
Fast Pose (DCN)ResNet50 - dcnYOLOv3256x192Symmetric Integral57.716.70 iter/sGoogle Baidu(code: dq9k)cfg
Fast PoseResNet50YOLOv3256x192Combined57.810.28 iter/sGoogle Baidu(code: 7a56)cfg
Fast Pose (DCN)ResNet50 - dcnYOLOv3256x192Combined58.210.22 iter/sGoogle Baidu(code: 99ee)cfg
Fast Pose (DUC)ResNet152YOLOv3256x192Symmetric Integral56.915.72 iter/sGoogle Baidu(code: jw3u)cfg

Notes

  • All of above models are trained only on COCO WholeBody dataset.
  • The APs are tested under COCO WholeBody's criterion, with flip test on.
  • The speed is tested on COCO val2017 on a single NVIDIA GeForce RTX 3090 gpu, with batch_size=64 in each iteration and offline yolov3 human detection results.

Notes

  • These models are strongly recommended because they are more accurate and flexible.
  • These models are trained with multi-domain knowledge distillation (MDKD, see our paper for more details).
  • The APs are tested under Halpe's criterion, with flip test on.
  • If you want to use the single hand model, you should give the rough bounding box of a single hand instead of that of a whole person.
  • The speed is tested on COCO val2017 on a single NVIDIA GeForce RTX 3090 gpu, with batch_size=64 in each iteration and offline yolov3 human detection results.

3D Human Pose & Shape Estimation

ModelBackboneInput SizePA-MPJPE (3DPW)PA-MPJPE (Human3.6M)DownloadConfig
HybrIKResNet34256x25645.336.3modelcfg