There are four categories in video list : org_video, label, start_frm, video_id
And I want to use extract_features.py for untrimmed video datasets.
Here is my questions.
1) Does "Clip" you used correspond to "snippet" meaning several successive frames?
2) As I understand, I should use different video_id for different clips in one video for feature extraction of untrimmed video. Right?
3) When using video_id, Is the start_frm meaning the target frame of label?
And that's why you use 1 frame clips in tutorial.
I have some questions about video list file for feature extraction. I saw your answer of previous issue where https://github.com/facebookresearch/VMZ/issues/11#issue-331330979, but I'm still not sure about my understanding.
There are four categories in video list : org_video, label, start_frm, video_id And I want to use extract_features.py for untrimmed video datasets. Here is my questions.
1) Does "Clip" you used correspond to "snippet" meaning several successive frames? 2) As I understand, I should use different video_id for different clips in one video for feature extraction of untrimmed video. Right? 3) When using video_id, Is the start_frm meaning the target frame of label? And that's why you use 1 frame clips in tutorial.
Thank you.