JVET-AF0152 AhG14: SADL update [F. Galpin, T. Dumas, P. Bordes, E. François (InterDigital)]

This contribution presents updates in the Small AdHoc Deep Learning (SADL) library. The update contains both some fixes and new features.

New layers

The following new layers were added (thanks to W. Bao for high quality MR):

• Grid-sample layer: float and integer support (e.g. support for motion compensation)
• Resize layer

Improvements and fixes

• Fix for slicing (support for LOP)
• Extend Conv2Dtranspose to new sizes
• Corrected leaky ReLU saturation for |alpha|>1
• Corrected sample programs