Toolbox for the ONERA Delta project.

This toolbox provides code sample developped in the project for various applications. It will be enriched as the project goes on.

The main Delta Toolbox code is in deltatb folder. Examples are in the workspace folder.

The Delta toolbox provides utility scripts for dense image machine learning with PyTorch.

• Dataset folder contains datasets classes in PyTorch format for loading multiple images and targets for dense prediction such as semantic segmentation.
• networks folder contains scripts for several networks, including UNet, SegNet...

The optical flow estimation using PWC-Net needs a custom correlation layer to be installed : cd deltatb/networks/correlation_package python ./setup.py install Code from github : NVIDIA/flownet2-pytorch, modified to work on pytorch 1.0, cuda 10

Semantic segmentation example is in the workspace:

• ISPRS 2D semantic segmentation