metrics about the test

[RitaFerreira03]

Hello
Is it possible to generate the metrics results that are output on train and validation, but for the test?

[glenn-jocher]

@RitaFerreira03 to evaluate a model on each dataset split:

python val.py --task train
python val.py --task val  # default
python val.py --task test

[RitaFerreira03]

I think I found out why! I think the error was in the yaml file which didn't have the path to the test dataset.
Can you please confirm if that makes sense?

Thank you once again!

[glenn-jocher]

@RitaFerreira03 above commands run val.py on your 3 datasets splits. Test set is optional.

yolov5/data/VisDrone.yaml

Lines 12 to 14 in a4fece8

	train: VisDrone2019-DET-train/images # train images (relative to 'path') 6471 images
	val: VisDrone2019-DET-val/images # val images (relative to 'path') 548 images
	test: VisDrone2019-DET-test-dev/images # test images (optional) 1610 images

[RitaFerreira03]

yes, that is what I have in the yaml file.
I just wanted to know if I can have metrics output when I run detect.py and how. The output are the test images with the objects detected, but that's it?

[glenn-jocher]

@RitaFerreira03 val.py produces metrics. detect.py simply runs inference.

[RitaFerreira03]

Thank you @glenn-jocher. Nice work with the algorithm!