Closed HYERI520 closed 3 years ago
I also face the same problem. I wanna hear a solution. Thank you.
I didn't really understand the question. We can just treat the model as a detector and a re-id feature extractor, every detection bbox is associated with a re-id feature. Even if there is no query image, we can still get the detection bbox and re-id feature.
Can you show me how to calculates single image mAP and saves or shows the topk images with the highest scores based on prediction results?
https://github.com/daodaofr/AlignPS/blob/master/tools/test_results.py
The test file shows how to calculate mAP. For visualization, you can sort the ranking results and draw the bounding box on the image.
Hi, Do you have any guideline to use final similarity score? I want to know, how does it work when there is no query person in just one image? Even if it is an incorrect result, since the person with the highest score is found?
Thank you.