search

alexlee-gk / video_prediction

Stochastic Adversarial Video Prediction
https://alexlee-gk.github.io/video_prediction/
MIT License
303 stars 65 forks source link

About the save_tf_record() #14

Closed happyday521 closed 5 years ago

happyday521 commented 5 years ago

Hi! Thanks for releasing the code! I am having some trouble preparing my own dataset. In function read_videos_and_save_tf_records（）: clipboard I think one sequence corresponds to the frames of one video file（for example 1.avi） ，and the function save_tf_record（） save these frames to a .tfrecord file（e.g 1.tfrecord）.

However, in the function save_tf_record（）: clipboard1 It seems one sequence corresponds to >= 1 video files. I am confused by it,

Can you help me? Thanks!

alexlee-gk commented 5 years ago

That's right, the function save_tf_record can save multiple sequences corresponding to multiple files. However, this function was always just called with a single sequence.

This turned out to be quite inefficient. I have updated the master branch with several improvements, one of which is in the preprocessing of the KTH dataset. In the new version, you should see that sequences_per_file=128 video sequences are saved in each tfrecord file.

Also, after you pull my changes, make sure to pip install -r requirements.txt again since there is one new dependency.

happyday521 commented 5 years ago

Get it. Thanks！