This is a re-release of the model_seg_rabbit_axon-myelin_bf model trained with nnunetv2. The previous one only contained the last checkpoints, which are outperformed by the "best checkpoints", as the latter gave the best validation Dice during training and look less overfitted. We also include the model trained on the full training set with the fold_all option, which is 5x lighter and faster than the ensembled model.
For more information, see #2 and axondeepseg/axondeepseg#771.