[TPAMI 2020] "Privacy-Preserving Deep Action Recognition: An Adversarial Learning Framework and A New Dataset" by Zhenyu Wu, Haotao Wang, Zhaowen Wang, Hailin Jin, and Zhangyang Wang
Is the CMAP metric from calling the average_precision_score() function on frame level?
Like if the 515 videos of PA-HMDB have in total 50k frames, then CMAP is from using average_precision_score(gt, output), where output is in the shape of [50k, 5]๏ผ
Thank you very much for creating this dataset. I am trying to replicate the results you reported in the TPAMI paper, and I would like to know how you split the videos from the PA-HMDB dataset? In order to compare my approach with yours, it is important to have the same train/test split. Would you please provide the train and test split files or explain how you split the dataset?