openpifpaf

We propose a new bottom-up method for multi-person 2D human pose estimation that is particularly well suited for urban mobility such as self-driving cars and delivery robots. The new method, PifPaf, uses a Part Intensity Field (PIF) to localize body parts and a Part Association Field (PAF) to associate body parts with each other to form full human poses. Our method outperforms previous methods at low resolution and in crowded, cluttered and occluded scenes thanks to (i) our new composite field PAF encoding fine-grained information and (ii) the choice of Laplace loss for regressions which incorporates a notion of uncertainty. Our architecture is based on a fully convolutional, single-shot, box-free design. We perform on par with the existing state-of-the-art bottom-up method on the standard COCO keypoint task and produce state-of-the-art results on a modified COCO keypoint task for the transportation domain.

@article{kreiss2019pifpaf,
  title = {PifPaf: Composite Fields for Human Pose Estimation},
  author = {Kreiss, Sven and Bertoni, Lorenzo and Alahi, Alexandre},
  journal = {CVPR},
  year = {2019}
}

Demo

Image credit: "Learning to surf" by fotologic which is licensed under CC-BY-2.0.

Created with:

python -m openpifpaf.predict \
  --checkpoint outputs/resnet101block5-pifs-pafs-edge401-l1-190131-083451.pkl \
  data-mscoco/images/val2017/000000081988.jpg -o docs/coco/ --show

Install

Python 3 is required. Python 2 is not supported.

pip install openpifpaf

# for development, install from source:
pip install numpy cython
pip install --editable '.[train,test]'

For a live demo, we recommend to try the openpifpafwebdemo project. Alternatively, openpifpaf.webcam provides a live demo as well. It requires OpenCV. To use a globally installed OpenCV from inside a virtual environment, create the virtualenv with the --system-site-packages option and verify that you can do import cv2.

Interfaces

• python -m openpifpaf.train --help
• python -m openpifpaf.eval_coco --help
• python -m openpifpaf.logs --help
• python -m openpifpaf.predict --help
• python -m openpifpaf.webcam --help

Example commands to try:

# live demo
MPLBACKEND=macosx python -m openpifpaf.webcam --scale 0.1 --source=0

# single image
python -m openpifpaf.predict my_image.jpg --show

Pre-trained Networks

Put these files into your outputs folder: Google Drive

Visualize logs:

python -m pifpaf.logs \
  outputs/resnet50-pif-paf-rsmooth0.5-181209-192001.pkl.log \
  outputs/resnet101-pif-paf-rsmooth0.5-181213-224234.pkl.log \
  outputs/resnet152-pif-paf-l1-181230-201001.pkl.log

Train

See datasets for setup instructions. See studies.ipynb for previous studies.

Train a model:

python -m openpifpaf.train \
  --lr=1e-3 \
  --epochs=75 \
  --lr-decay 60 70 \
  --batch-size=8 \
  --basenet=resnet50block5 \
  --headnets pif paf \
  --square-edge=401 \
  --regression-loss=laplace \
  --lambdas 30 2 2 50 3 3 \
  --freeze-base=1

You can refine an existing model with the --checkpoint option.

To produce evaluations at every epoch, check the directory for new snapshots every 5 minutes:

while true; do \
  CUDA_VISIBLE_DEVICES=0 find outputs/ -name "resnet101block5-pif-paf-l1-190109-113346.pkl.epoch???" -exec \
    python -m openpifpaf.eval_coco --checkpoint {} -n 500 --long-edge=641 --skip-existing \; \
  ; \
  sleep 300; \
done

Person Skeletons

COCO / kinematic tree / dense:

Created with python -m openpifpaf.data.

Video

Processing a video frame by frame from video.avi to video-pose.mp4 using ffmpeg:

ffmpeg -i video.avi -qscale:v 2 -vf scale=641:-1 -f image2 video-%05d.jpg
python -m openpifpaf.predict --checkpoint outputs/resnet101block5-pifs-pafs-edge401-l1-190213-100439.pkl video-*0.jpg
ffmpeg -framerate 24 -pattern_type glob -i 'video-*.jpg.skeleton.png' -vf scale=640:-1 -c:v libx264 -pix_fmt yuv420p video-pose.mp4

Evaluations

See evaluation logs for a long list. This result was produced with python -m openpifpaf.eval_coco --checkpoint outputs/resnet101block5-pif-paf-edge401-190313-100107.pkl --long-edge=641 --loader-workers=8:

Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets= 20 ] = 0.662
Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets= 20 ] = 0.872
Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets= 20 ] = 0.724
Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets= 20 ] = 0.623
Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets= 20 ] = 0.721
Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 20 ] = 0.712
Average Recall     (AR) @[ IoU=0.50      | area=   all | maxDets= 20 ] = 0.895
Average Recall     (AR) @[ IoU=0.75      | area=   all | maxDets= 20 ] = 0.768
Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets= 20 ] = 0.660
Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets= 20 ] = 0.785