search

Went-Liang / UnSniffer

[CVPR 2023] Official Pytorch code for Unknown Sniffer for Object Detection: Don’t Turn a Blind Eye to Unknown Objects
https://arxiv.org/pdf/2303.13769v1.pdf
Apache License 2.0
67 stars 9 forks source link

Question about how to get the OW-DETR's results under the evaluation of this paper #3

Open harrylin-hyl opened 1 year ago

harrylin-hyl commented 1 year ago

Wonderful job! I notice you evaluate the performance of OW-DETR under the evaluation of this paper. I want to know how can I get the results.

Went-Liang commented 1 year ago

Thank you for your interest in our model!

You can rewrite the evaluate function so that it can save the json file in the same format as our output. Then you need to use these files to evaluate the performance of OW-DETR.

harrylin-hyl commented 1 year ago

Thanks for your quick reply! When I implemented the results of OW-DETR, I noticed that the mixed_OOD file has 704 images instead of 897 in the paper, but mix_ID file does have 897 images. Does this mean that some of these images do not have OOD objects?

Went-Liang commented 1 year ago

Regarding your observation, you are correct. The reason for this difference is that during the evaluation of OOD objects, we excluded the images from the ground truth that do not contain any OOD objects.