The Unet is an expensive architecture, so it takes awhile on the CPU. You could try lowering the image resolution some or use some of the speed ups I mention here and see: Speeding Up fastai2 Inference - And A Few Things Learned
1 Like