search

wangxiang1230 / OadTR

Code for our ICCV 2021 Paper "OadTR: Online Action Detection with Transformers".
MIT License
87 stars 12 forks source link

fps of the original video #18

Closed Echo0125 closed 2 years ago

Echo0125 commented 2 years ago

Hi Thank you for sharing the dataset. But I met some problems when I used the dataset. As you said in the previous issue, the fps of the origin video is 24fps, so the actual fps of the feature is 4, but I calculate the frames and the seconds of the video found that the fps of the feature is 5... If there is any mistakes?

wangxiang1230 commented 2 years ago

Hi Thank you for sharing the dataset. But I met some problems when I used the dataset. As you said in the previous issue, the fps of the origin video is 24fps, so the actual fps of the feature is 4, but I calculate the frames and the seconds of the video found that the fps of the feature is 5... If there is any mistakes?

Thanks for your attention to OadTR. We extract video frames at the original frame rate of the video, and extract a feature every 6 frames. Each video may vary from 24FPS to 30FPS. In the experiment, we also found that the frame rate has little effect on the final result (e.g., 24FPS vs. 30FPS), a very simple way to verify is that you can linearly interpolate all the features to the 24FPS video input, and then compare their performance. It should be noted that the label shape also needs to be resized accordingly.