google-deepmind / kinetics-i3d

Convolutional neural network model for video classification trained on the Kinetics dataset.
Apache License 2.0
1.74k stars 461 forks source link

extract features #100

Open Serendipity-LC opened 4 years ago

Serendipity-LC commented 4 years ago

Hi!How to use I3D to extract features from other datasets? If n represents the frames of a video, How can I extract features of size n*1024 with rgb_videos as the input?

Nandini2801 commented 3 years ago

Hi @chenjiakun I also had similar ques. Have you got any progress for the above problem?

Tortoise17 commented 3 years ago

@Nandini2801 This is pytorch version Kinetics-PyTorch which is working. But still I need to know how to extract features from second last layer.

Nandini2801 commented 3 years ago

@Tortoise17 how to make this code to extract features from other datasets rather than kinetic?

Tortoise17 commented 3 years ago

Sorry, now in above repl is the code.

Nandini2801 commented 3 years ago

Thanks for the reply