when train_backbone，got a bad performance!

DonaldRR / SimpleNet

MIT License

402 stars 59 forks source link

when train_backbone，got a bad performance! #47

Closed PHaiJun closed 8 months ago

PHaiJun commented 8 months ago

why??

INFO:main:instance_auroc: 0.682 INFO:main:full_pixel_auroc: 0.661 INFO:main:anomaly_pixel_auroc: 0.007 INFO:main:

INFO:utils:instance_auroc: 0.682 INFO:utils:full_pixel_auroc: 0.661 INFO:utils:anomaly_pixel_auroc: 0.007

PHaiJun commented 8 months ago

@DonaldRR

DonaldRR commented 8 months ago

The FREEZING backbone is one of the key in this work. The pretrained backbone models the "normal" features distribution. When it's parameters are changing over time, such distribution will be changing too.

DonaldRR commented 8 months ago

The essence of the training is not always Decreasing the Loss. The freezing backbone makes the features from backbone are meaningful,

sdracula commented 7 months ago

@DonaldRR

excuse me, how could you get the good performence by this resposiry