Notebooks to: train CamVid-Tiramisu dataset to 95% accuracy (sota), convert CamVid dataset to Tiramisu paper version, and other training.
I did this on Windows 10 with the fast.ai v1 library: https://github.com/fastai/fastai
CamVid Datset: http://mi.eng.cam.ac.uk/research/projects/VideoRec/CamVid/
The One Hundred Layers Tiramisu Fully Convolutional DenseNets for Semantic Segmentation paper: https://arxiv.org/pdf/1611.09326.pdf