Chuming Lin
Chuming Lin
@menghuaa , in activitynet 1.3, we follow GTAD/BMN methods and use the class predict results of the CUHK untrimmednet as the action class scores.
你好，我们当时的测试环境是 V100 + cuda 10.2 + pytorch 1.4，但是我们这边没有RTX 3090，所以没有办法帮你在该硬件上测试，你可以先确认一下你安装的pytorch版本是不是GPU版本
@bleakie 可以参考一下THUMOS14的分类训练与测试，然后套在anet上即可
@lurchycc good issue! In experiments, we directly took the THUMOS GT file from GTAD codes (https://github.com/frostinassiky/gtad/blob/master/evaluation/thumos_gt.json). To compare to GTAD, we did not change the json file. You can update...
@menghuaa, I think you means that we remove the 'Ambiguous' class label. These data label is useless to training and testing for action detection task. GTAD is an action proposal...
@menghuaa you can observe the gt file for ambiguous class to action proposal task. If ambiguous proposals are included in gt file, you should not remove them in training and...
@menghuaa no, I have never seen it.
@fanqing-1116 , did you try use `7za x -r train_val_112.7z` and check the number of mp4 files is 14950?
@czhaneva you can refer to the answer of the issue #4 .
@shubhamagarwal92, thanks for your attention! If the duration of the action is short and the video length is very long, you can refer the THUMOS data processing. If the range...