I found a new tool makesense who is trying to do the same thing that you're already doing. Probably with some new thoughts in mind. I recently asked the author how that tool is doing differently: SkalskiP/make-sense#23

The author responded that we need multiple clicks while labeling in imglab. Though I didn't understand it well as I can control most of the things with

When publishing a dataset and when training a model (to avoid bias) it could be useful to remove sensitive information like car plate numbers, names and other personal info on medical images, faces etc. Another variant - is removal of background (noise) from images for anomaly detection tasks. One more - removal of reflections on the image.

Thanks for your nice work and congratulations on your good results!

I have several questions.

• Will your model extended to Parallel (distributed data-parallel) in the future.
• Why don't you try to use deeplabv3+, will it lead to a better result?

Best.