jdai-cv/pgpt

Implementation of ‘Pose-Guided Tracking-by-Detection: Robust Multi-Person Pose Tracking’ [TMM 2020]

pgpt's Issues

detection file

could you provide the detection file for validation dataset, please?

demo_detection.json file has different format as it should in skeleton_visulize.py

Hi, thank for sharing your work,

I try to run the demo.sh but it seems that your demo_detection.json is in another format and the skeleton_visulize.py can not read it.

I try to change the python script, I can change the first section to read the correct frame in the correct folder, but I'm stuck with the keypoint list old_point_list = anno['annopoints'][0]['point']. Since the actual format of the json file is (as I believe): { "dataset_i": {"images": [ [x_min, x_max, y_min, y_max, track_id] ] }, }, where I can get the pose detection?

bests,
looninho

I‘cant find embedding model here. Could you provide the model please?

Why there are only 15 key points not 17

in inference.py
For posetrack dataset, the output keypoints is as follow only 15 keypoints，but in posetrack dataset has 17 keypoint。
Left ear and Right ear lost.