What returns Keras Conv2DLSTM module

Hello everyone,

I would like to know if keras’s Conv2DLSTM module returns next N frames (in t, t+1, t+2…t+N-1) time or sequences in (t-N,…t-1) time? I’m trying to predict next N frames.

Also I what to know if I set it to return_sequences=False does it return me next frame in sequence?

Please provide some source for your answers.

Hi @Tomas_Cernacek, By assigning return_sequences=False does not return the next frame in sequence. The return_sequences argument is used to whether the layer returns the entire output sequence(including the hidden states of the Lstm) for each input or just the final output.

please refer to this document to know the output of Conv2DLSTM and how it depends on value assigned to the return_sequences argument. Thank You.