Segmenter training data and trained models

From the tutorials on segmentation I understand that there is a pipeline to be followed. First classic segmentation is done followed by DL segmentation using the good classic results as ground truth (curation).
It makes sense that after a while there would be a repository of ground truth data to use for training w/o the need for each individual to do the manual classic-> curation part. In addition it also makes sense that pretrained DL models/weights would be available to use and therefore only fine tuning on custom data is needed.
Is there such data/models available?


Hi oded! Thanks for your interest in our tools and data. One of our recent papers releases a dataset with images and ground truths that could be used for training a segmentation model. Model weight release is something we’re thinking about, so it’s good to hear that it would be of interest to the community!


@oded The weights for some of these segmentation models are also available in our Segmenter Model Zoo (details here, repo here)