The pretrained model you provided perform better than DPL(your method) in domain1

cchen-cc / SFDA-DPL

[MICCAI'21] Source-Free domain adaptive fundus image segmentation with denoised pseudo-labeling

51 stars 9 forks source link

The pretrained model you provided perform better than DPL(your method) in domain1 #7

Open chenslcool opened 1 year ago

chenslcool commented 1 year ago

I got OC=0.848 and OD=0.959 on domain1 with your pretrained model. However, after adaptation with DPL, OC=0.8222,OD=0.9446, I repeat for 3 times and got same result. Maybe your method cannot perform well on domain1?