TIO-IKIM / CellViT

CellViT: Vision Transformers for Precise Cell Segmentation and Classification
https://doi.org/10.1016/j.media.2024.103143
Other
217 stars 33 forks source link

Training is currently broken #16

Closed FabianHoerst closed 1 year ago

FabianHoerst commented 1 year ago

Describe the bug Currently the training script contains bugs. We compared the script provided here with the original training scripts used for the publication and found some bugs. We are currently fixing the bugs.

Inference Scripts are not affected and can be used without any limitation. This only affects users that are trying to fine-tune the networks on new data or try to reproduce the results given in the paper.