How can I build a benchmark for Image Segmentation?

[KirigiriSuzumiya]

I don't know whether it's possible to train an Image Segmentation model on avalanche, as I find that most of the benchmarks are build on classification tasks. For instance, how can I train a U-Net with a benchmark whose output layer is not a single class, but like [w,h]?

[Ali-Olliek]

Hello.

Did you find a way to achieve this?

[KirigiriSuzumiya]

yes, you can take a look at my code here. Most of the work is on building the benchmark

[Ali-Olliek]

Thank you!