mala-lab / InCTRL

Official implementation of CVPR'24 paper 'Toward Generalist Anomaly Detection via In-context Residual Learning with Few-shot Sample Prompts'.
Apache License 2.0
83 stars 12 forks source link

Can you provide code for visual inspection #8

Closed czr1030 closed 3 months ago

czr1030 commented 3 months ago

Thank you for your work. Can you provide the code for visual detection

Diana1026 commented 3 months ago

Hello, may I know what the "visual detection" means?

czr1030 commented 3 months ago

Hello, may I know what the "visual detection" means?

Generate a heatmap of abnormal locations

Diana1026 commented 3 months ago

In our paper, after obtaining the holistic in-context residual map M, we use sns.heatmap(M, linewidth=0.5, vmin=0, vmax=1, cmap="YlGnBu") to generate the heatmap. By the way, this heatmap is not intended for the anomaly segmentation task, which is not involved in InCTRL.