search

astorfi / lip-reading-deeplearning

:unlock: Lip Reading - Cross Audio-Visual Recognition using 3D Architectures
Apache License 2.0
1.84k stars 323 forks source link

Number of filters #22

Closed ir0nt0ad closed 4 years ago

ir0nt0ad commented 5 years ago

I may be mistaken, but shouldn't we be creating 128, not 64 filters here? The comment below says shape=(?, 9, 128); we also had 128 in the previous layer.

https://github.com/astorfi/lip-reading-deeplearning/blob/6fe954015b796f70daa137c5c266f230a37ab6b2/code/training_evaluation/nets/lipread_speech.py#L138 https://github.com/astorfi/lip-reading-deeplearning/blob/6fe954015b796f70daa137c5c266f230a37ab6b2/code/training_evaluation/nets/lipread_speech.py#L145

astorfi commented 4 years ago

That does not make a huge difference.