Cant achieve Metrics as mentioned in the paper for consep and cpm_17 dataset #1
Description
Hi!
I have tried to train CompSegNet using same repo and same dataset provided by the Author but I am not successful in achieving the metrics claimed in paper. Even when i am trying to train model on cpm-17 and consep the validation loss is not decreasing and remain above 0.7. If anyone know how to properly implement it kindly let me know. It worked on monuseg somehow. but for other 2 datasets its very difficult.