mcd8604/rvsc

RVSC - Retinal Vessel Segmentation and Classification

Caffe models can be trained, tested, and evaluated as described in the respective README files:

• 4-Class Model
• 3-Class Model
• 2-Class Model

rvsc's Issues

how to finetune

Thanks for your great project.

When I try to use 3_class model with pretrained weight on RITE dataset, the accurate result is obtained. However, the classifiction of small artery and vein is not very well, I want to finetune the model. After 20000 iterations based on your pretrained weight, the segmentation is almost white, which is much worse than yours. Can you explain how to finetune your model?