Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Visualization settings with MODEL.ROI_HEADS.SOFT_NMS_ENABLED

See original GitHub issue

Why is MODEL.ROI_HEADS.SOFT_NMS_ENABLED True enabled in test_transfer_learning.sh and not in visualize_transfer_learning.sh. And when it is enabled in visualize_transfer_learning.sh the generated images sometimes output accuracies greater than 100% for detected bounding box classes. Why is that?

Issue Analytics

State:
Created a year ago
Comments:5

Top GitHub Comments

1reaction

YiwuZhongcommented, Jul 23, 2022

In regular NMS setting, setting MODEL.CLIP.VIS True can ensure visualized scores to be always <100%.

In soft NMS setting, setting MODEL.CLIP.VIS True cannot make the visualized scores to be <100% (not implemented yet). But you can either set MODEL.CLIP.MULTIPLY_RPN_SCORE False or disable soft-version NMS to visualize scores <100%.

I would recommend simply disabling soft NMS (minimal effects on detection results) but keep original setting of MODEL.CLIP.MULTIPLY_RPN_SCORE (more effects on detection results).

0reactions

Jawingcommented, Jul 23, 2022