Details of finetuning UCF101 and HMDB51

[icyzhang0923]

In the TABLE 1 of the paper reports the TSN and TSM results on UCF101 and HMDB51 dataset, however, could you give more details about the experiments settings: 1、Did you follow TSN to process the video data to extract rgb frames? What the scale of the rgb frames are? 2、The pre-trained model you use is dense sampling or uniform sampling? 3、The results on UCF101 and HMDB51 are reported on only split 1 or the 3 splits average? 4、The results on UCF101 and HMDB51 are reported on only rgb input or fusion the flow?

[jyGuan]

Yes! I have the same problem.