Generate randomly defects on positive samples and restore them with GAN.
-
It is built from the Pix2Pix respository.
-
There are three different models, consisting of autoencoder, segment model(Unet), and GAN (Pix2Pix).
-
It is inspire by the paper "A Surface Defect Detection Method Based on Positive Samples".
-
We Got the 5th place in the preliminary contest of National campus machine vison application competition.