search

gaobb / OneNIP

[ECCV 2024] Learning to Detect Multi-class Anomalies with Just One Normal Image Prompt
https://csgaobb.github.io/Pub_files/ECCV2024_OneNIP_CR_Full_0725_Mobile.pdf
24 stars 1 forks source link

Visualize Reconstructed Features #4

Open k1L1t0 opened 1 month ago

k1L1t0 commented 1 month ago

Hellow, thanks for your nice code. I've just finished a reproduction of UniAD and want to make a comparison with OneNIP. How can I train the vis_decoder and visualize the reconstructed features like UniAD?

gaobb commented 1 month ago

Hi, @k1L1t0

We did not study the visualization of reconstruction features in our OneNIP paper, so this repo does not provide the corresponding implementation code.

However, this is worth exploring further. You can implement it following the code (./tools/train_vis_decoder.py) of UniAD.