AST This is unofficial implementation of "Asymmetric student-teacher networks for industrial anomaly detection" Write your data directory in config.py set config.py train_teacher.py train_student.py eval.py Result MVtecAD Image-level AUROC mean max Paper leather 1 1 1 zipper 0.991 0.977 0.991 metal_nut 0.989 0.996 0.985 wood 0.988 0.992 1 pill 0.992 0.964 0.991 transistor 0.990 0.987 0.993 grid 0.990 0.999 0.991 tile 0.999 0.996 1 capsule 0.992 0.971 0.997 hazelnut 0.998 0.997 1 toothbrush 0.961 0.864 0.966 screw 0.993 0.944 0.997 carpet 0.972 0.972 0.975 bottle 0.998 0.994 1 cable 0.992 0.939 0.985 MVtecAD-3D Image-level AUROC mean max foam 0.864375 0.89875 tire 0.68092 0.665287 peach 0.821843 0.993832 cable_gland 0.910783 0.958402 carrot 0.8656 0.992985 rope 0.927989 0.880435 potato 0.666502 0.935277 cookie 0.93516 0.992718 bagel 0.855888 0.910124 dowel 0.984837 0.95858 AVerage 0.85139 0.918639