Not sure whether this should be merged but at least wanted to share.
Regenerate SegEM training data without erosion and store to wkwrap instead of .mat for easier visualization. Data can be inspected with webKnossos here.
Remaining ToDo could be to "globalize" the segmentation ids. Currently all 100^3 training cubes are numbered 1...N which makes the "highlight hovered cells" feature in wK a bit unintuitive.
See here for the list of training regions in json format:
/tmpscratch/webknossos/Connectomics_Department/2012-09-28_ex145_07x2_segEM_training_data_no_erosion/segmentation/1/stacks.json
and the webknossos dataset in that folder for the segmentation and raw data.
Not sure whether this should be merged but at least wanted to share. Regenerate SegEM training data without erosion and store to wkwrap instead of .mat for easier visualization. Data can be inspected with webKnossos here.
Remaining ToDo could be to "globalize" the segmentation ids. Currently all 100^3 training cubes are numbered 1...N which makes the "highlight hovered cells" feature in wK a bit unintuitive.
See here for the list of training regions in json format:
/tmpscratch/webknossos/Connectomics_Department/2012-09-28_ex145_07x2_segEM_training_data_no_erosion/segmentation/1/stacks.jsonand the webknossos dataset in that folder for the segmentation and raw data.