HRNet / HigherHRNet-Human-Pose-Estimation

This is an official implementation of our CVPR 2020 paper "HigherHRNet: Scale-Aware Representation Learning for Bottom-Up Human Pose Estimation" (https://arxiv.org/abs/1908.10357)
MIT License
1.35k stars 272 forks source link

Keypoints output result json format problem #45

Open mytk2012 opened 4 years ago

mytk2012 commented 4 years ago

Keypoints output result json referred to image_id which is repeatable seemes not to be effective,when i try valid experiment using my own dataset.Two different pictures but the image_ids of them are the same.Can you explain this?I have read your paper often,but I cannot figure it out.