You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hey,
Thank you very much for your video generator!
I am now trying to train a LSTM with image sequences but as I want to train for a future GAN implementation I have no classes. How would I be able to apply your generator with no classes?
Would be grateful if you could help.
Thanks a lot
Cate
The text was updated successfully, but these errors were encountered:
I already tried (without good results for now). The idea is to make the "generator" (I mean the model part, not the VideoFrameGenerator) to output a shape of (B, N, W ,H , C) where B is the batch size (None ?) and N the number of frame.
Then use VideoFrameGenerator (or SlidingFrameGenerator) dropping class:
# index 0 is the batches of sequeces, index 1 is classes, you can ignore this indexbach=next(your_generator)[0]
I think the idea can be very interesting without using data augmentation (or not destructive).
I will try to give more details in my Medium account (my blog is french, my Medium is english)
Yes! Actually I am trying to develop a many-to-one model so the output of the generator should only be one next frame.
Will work on it and share if I manage it successfully!
Thanks a lot for your reply and Ill wait for your Medium post.
Hey,
Thank you very much for your video generator!
I am now trying to train a LSTM with image sequences but as I want to train for a future GAN implementation I have no classes. How would I be able to apply your generator with no classes?
Would be grateful if you could help.
Thanks a lot
Cate
The text was updated successfully, but these errors were encountered: