search

zhiyuanyou / UniAD

[NeurIPS 2022 Spotlight] A Unified Model for Multi-class Anomaly Detection
Apache License 2.0
250 stars 28 forks source link

Test.json file #36

Closed ew-kim closed 1 month ago

ew-kim commented 1 month ago

Hi! Thank you for your great work. I have a question. I want to evaluate uniad with a new dataset, but I encountered an issue when I modified the test.json file. If one test image has more than one mask image, can I know how to modify it? In this case, do we only use one mask or do we have to use another method?

zhiyuanyou commented 1 month ago

Hello, thanks for your interests.

However, I wonder why one image has more than one mask image?

Is it because there are multiple ground truth?

ew-kim commented 1 month ago

Yes, there are multiple ground_truths. For example, for the MVTec-LOCO dataset, there are multiple ground_truths (logical_anomalies) for pushpins.

zhiyuanyou commented 1 month ago

Sorry that I am not familiar with MVTec-LOCO dataset. UniAD can only produce a single prediction. Maybe evaluating with every ground-truth mask is a possible solution.

ew-kim commented 1 month ago

Thank you for your thoughtful consideration and kind response to my question.