AlphaPose

AlphaPose is an accurate multi-person pose estimator, which is the first open-source system that achieves 70+ mAP (75 mAP) on COCO dataset and 80+ mAP (82.1 mAP) on MPII dataset. To match poses that correspond to the same person across frames, we also provide an efficient online pose tracker called Pose Flow.

Halpe 136 keypoints YouTube link

System Used

This system was tested on Ubuntu 20.04 with Nvidia RTX 3060 GPU, Cuda 11.3 and Nvidia Driver 535.

Setup

Clone the repo:

git clone https://github.com/ArghyaChatterjee/AlphaPose.git

Create a virtual environment:

cd AlphaPose/
python3 -m venv alpha_pose_venv
source alpha_pose_venv/bin/activate
pip3 install --upgrade pip
pip3 install torch==1.12.1+cu113 torchvision==0.13.1+cu113 torchaudio==0.12.1 --extra-index-url https://download.pytorch.org/whl/cu113
pip3 install -r requirements.txt
echo "export PYTHONPATH=$(pwd):\${PYTHONPATH}" >> ~/.bashrc
source ~/.bashrc
sudo apt install libyaml-dev
python3 setup.py build develop

Install Pytorch 3D (Optional):

conda install -c fvcore -c iopath -c conda-forge fvcore iopath
conda install -c bottler nvidiacub
pip3 install git+ssh://git@github.com/facebookresearch/pytorch3d.git@stable

Download Pretrained Models

1. Download the object detection model manually: yolov3-spp.weights(Google Drive | Baidu pan). Place it into detector/yolo/data.
2. Or, if you want to use YOLOX as the detector, you can download the weights here, and place them into detector/yolox/data. We recommend yolox-l and yolox-x.
3. Download the pose models and place them into pretrained_models. All models and details are available in the Model Zoo.

Tracking

Please read trackers/README.md for details.

Halpe Pose

Inference Demo (Webcam)

• Halpe 136 Keypoints Combined Pose Estimation

cd AlphaPose/ 
python3 scripts/alpha_pose_webcam_demo.py --cfg /home/arghya/alpha_pose/AlphaPose/configs/halpe_136/resnet/256x192_res50_lr1e-3_2x-dcn-combined.yaml --checkpoint /home/arghya/alpha_pose/AlphaPose/pretrained_models/halpe136_fast50_dcn_combined_256x192_10handweight.pth --outdir examples/res --vis --webcam 0

• Halpe 136 Keypoints DCN Regression Pose Estimation

cd AlphaPose/ 
python3 scripts/alpha_pose_webcam_demo.py --cfg /home/arghya/alpha_pose/AlphaPose/configs/halpe_136/resnet/256x192_res50_lr1e-3_2x-dcn-regression.yaml --checkpoint /home/arghya/alpha_pose/AlphaPose/pretrained_models/halpe136_fast50_dcn_regression_256x192.pth --outdir examples/res --vis --webcam 0

• Halpe 136 DUC Keypoints Regression Pose Estimation

cd AlphaPose/ 
python3 scripts/alpha_pose_webcam_demo.py --cfg /home/arghya/alpha_pose/AlphaPose/configs/halpe_136/resnet/configs/halpe_136/resnet/256x192_res152_lr1e-3_1x-duc.yaml --checkpoint /home/arghya/alpha_pose/AlphaPose/pretrained_models/halpe136_fast152_duc_regression_256x192.pth --outdir examples/res --vis --webcam 0

SMPL Pose

Colab Demo

colab example

Inference Demo

cd AlphaPose/
./scripts/inference.sh ${CONFIG} ${CHECKPOINT} ${VIDEO_NAME} # ${OUTPUT_DIR}, optional

Inference SMPL (Download the SMPL model basicModel_neutral_lbs_10_207_0_v1.0.0.pkl from here and put it in model_files/).

./scripts/inference_3d.sh ./configs/smpl/256x192_adam_lr1e-3-res34_smpl_24_3d_base_2x_mix.yaml ${CHECKPOINT} ${VIDEO_NAME} # ${OUTPUT_DIR}, optional

For high level API, please refer to ./scripts/demo_api.py. To enable tracking, please refer to this page.

Training

Train from scratch:

cd AlphaPose/
./scripts/train.sh ${CONFIG} ${EXP_ID}

Train FastPose on mscoco dataset.

cd AlphaPose/
./scripts/train.sh ./configs/coco/resnet/256x192_res50_lr1e-3_1x.yaml exp_fastpose

Validation

Validate your model on MSCOCO val2017:

cd AlphaPose/
./scripts/validate.sh ${CONFIG} ${CHECKPOINT}

Fast Pose

Inference Demo (Image Files)

If you want to use FastPose model + YOLOv3 SPP as the detector:

cd AlphaPose/
python3 scripts/demo_inference.py --cfg configs/coco/resnet/256x192_res50_lr1e-3_1x.yaml --checkpoint pretrained_models/fast_res50_256x192.pth --indir examples/demo/

If you want to use FastPose model + yolox-x as the detector:

python3 scripts/demo_inference.py --detector yolox-x --cfg configs/coco/resnet/256x192_res50_lr1e-3_1x.yaml --checkpoint pretrained_models/fast_res50_256x192.pth --indir examples/demo/

Inference Demo (Video Files)

If you want to use FastPose model + YOLOv3 SPP as the detector:

./scripts/inference.sh configs/coco/resnet/256x192_res50_lr1e-3_1x.yaml pretrained_models/fast_res50_256x192.pth ${VIDEO_NAME}

Crowd Pose

Please read docs/CrowdPose.md for details.

Detailed Demo

More detailed inference options and examples, please refer to GETTING_STARTED.md