Михаил

Thanks for the tests and the metrics! In fact, the error is still there, I can show you an example of what it looks like  As you can see,...

I hope these are the last fixes in this function))) https://github.com/MiXaiLL76/faster_coco_eval/commit/71b245df2ccd5b8af9332cf5ec438a2ab9174f85 Please, friend, do your tests again, compare the AP results between the libraries. I ran opencv vs maskapi and...

 The error with the borders looked something like this

Yes, errors in 10^-17 are usually an error in translating floating point numbers from C++ to python. It's not critical everywhere, I think it's a win, thanks for participating in...

I have read the CLA Document and I sign the CLA

> @MiXaiLL76 Is this package faster than the pycocotools meanwhile getting the same mAP values as pycocotools? Could you also post the mAP output from faster-coco-eval vs pycocotools? Thanks! Yes,...

> @MiXaiLL76 There's no output within this PR using faster-coco-eval package. > > ```shell > yolo detect val model=yolo11n.pt eval_backend=faster_coco_eval > ``` > Hi @Laughing-q ! It is true, my...

> @MiXaiLL76 Based on the speed up and it also remains the same results I think we should directly use faster-coco-eval instead of providing an option to select pycocotools or...

@Laughing-q @ambitious-octopus I am also generally for preserving variability in validation) pycocotools is the standard for original validation, but my library is faster and has more functionality. (though it is...

> @MiXaiLL76 After discussed with @ambitious-octopus we agreed with checking if the faster-coco-eval is available in the local env, and if it's available then use is as backend and if...