https://github.com/DonaldRR/SimpleNet

效果一般，但是真的简单??

1. Feature adaptor: a single fully connected layer
   1. input: a feature at location (h, w)
   2. fig 4, tab 3.
2. The Discriminator Dψ works as a normality scorer, estimating the normality at each location (h, w) directly.
   1. a 2-layer MLP
   2. no position info as input?
   3. 应该依赖多尺度feature，

SimpleNet: A Simple Network for Image Anomaly Detection and Localization, cvpr23