Closed paopao-long closed 2 years ago
Hi Asha, I'm sorry to bother you!I try to use ToDayGAN for training and testing CULane.But I got very strange results. I don't know why. The real night image is : Night image translated to day:
Hi,
this is not an unimaginable scenario. Could be you beed more data to train.
I would also try adding a —lambda-forward loss modifier option of something small like 0.2 or so.
Hi Asha, I'm sorry to bother you!I try to use ToDayGAN for training and testing CULane.But I got very strange results. I don't know why. The real night image is : Night image translated to day: