Without update the weights of EfficientNet during training

[gaowq2017]

Hi thanks for sharing the code, i have a question as follow: The weights of efficient net haven't update during training. Directly extracting feature using efficient net of pretrained weights is OK? or others? Thank you very much

[ghost]

Train the FLOW, not the FE.

[marco-rudolph]

We use pretrained weights from ImageNet and do not train them further. These are suitable for detecting anomalies. It does not really make sense to keep training these weights, because in this way features are learned whose density is good to model, but not necessarily helpful for anomaly detection.