This release brings incredible new features and improvements from the
community accumulated over the last months. We recommend to upgrade.
The pre-built docker images are the recommended way of using robosat:
Changes
-
rs train: state of the art losses and metrics. Lovasz loss as default,
many many more small features and fixes in training and related tools.
Thanks https://github.com/ocourtin -
rs extract: fully automatated road training dataset creation
Thanks https://github.com/DragonEmperorG -
rs extract: batch feature extraction for datasets too big for memory
Thanks http://github.com/daniel-j-h -
rs rasterize: batch rasterization for datasets too big for memory
Thanks http://github.com/daniel-j-h -
Infrastructure: improved docker images, pre-trained weights in images,
upgrades to CUDA 10.1, cudnn 7, and pytorch 1.1.
Thanks http://github.com/daniel-j-h