mittagessen
mittagessen
This is an implementation for unsupervised pretraining of recognition model weights based on an image inpainting surrogate task that aims to reconstruct randomly sampled masked patches from the initial convolutional...
Add a method to `lib.vgsl.TorchVGSLModel` to validate a (partial) spec, feeding a 'dummy' line/image into it depending on the input specification. This would allow aborting training before loading the training...
What it says on the tin.
This PR improves validation metrics for segmentation training by limiting the pixel metrics to regions, vectorizing baselines, and matching them to ground truth baselines.
The current API is rather specific to particular implementations of methods and keeps us from easily integrating methods that are not one to one replacements with similar semantics as current...
The library got relicensed under AGPL which is incompatible with Apache. This PR implements augmentation using torchvision primitives.