Main goal is to traing CNN-ish models to segment images. But to get to a trained model, many techniques need training masks. Perhaps various existing non-CNN image segmenting techniques can be used to generate masks. Maybe make multiple masks in parallel and then see if they agree.
Main goal is to traing CNN-ish models to segment images. But to get to a trained model, many techniques need training masks. Perhaps various existing non-CNN image segmenting techniques can be used to generate masks. Maybe make multiple masks in parallel and then see if they agree.
E.g. Random Walker segmentation in Python