Hiromu Taketsugu | 武次 広夢

Thanks for sharing! I think that is because Posetrack21 and JRDB have different skeleton definitions. When you fine-tune with a different keypoint structure, directly fine-tuning all weights doesn't work well....

Yes, that's correct, or as in the case of our paper, you can just run jrdbpose_train.py without a pretrained model and then use the trained model in run_active_learning.sh.

Thanks, there might be two reasons for this: 1. Please ensure that your training setup follows the procedure described in Section 5.1 (HP Estimator). 2. AP can vary significantly across...

Yes, your training process is correct. Could you share the average performance on test videos? It might also be helpful to confirm that the released pre-trained model works correctly ([email protected]...

Thank you for sending the log files. Normally, as the OKS threshold increases, the evaluation becomes stricter and AP values should decrease, but your results don't. Therefore, there may be...

Thanks for pointing this out! In Activelearning.py, since the fine-tuning and evaluation of a pose estimator is completed within the ATL process itself, we didn’t need to save the final...