Realtime_pose

Realtime 2d pose using camera.

Pose estimator: HRNet

Environment

python >= 3.6

pytorch >= 1.0

Anaconda3

run this create virtural env:

conda env create -f env_info_file.yml

Usage

python tools/hrnet_2d_realtime.py

Use with UI

you need PyQt

run this:

python tools/hrnet_2d_realtime_UI.py

Pre-trained model

• HRNet:

1. load pose model in

   joints_detectors/hrnet/models/pytorch/pose_coco/

2. load yolov3 in

   joints_detectors/hrnet/lib/detector/yolo/

   or download by

   wget https://pjreddie.com/media/files/yolov3.weights

• AlphaPose:

1. load due_se in

   joints_detectors/Alphapose/models/sppe/

2. load yolov3 (same to the yolov3 file in HRNet dir) in

   joints_detectors/Alphapose/models/yolo/

Reference

Modified from https://github.com/lxy5513/videopose