Getting AP metrics from darknet (AP@0.50:0.95 & more)

[pabsan-0]

Hello, I am trying to obtain a kind-of-normalized AP table for an academic project I'm doing in which I am training various object detectors, something similar to the following table (which was produced with efficientDet) with darknet:

 Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.304
 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.512
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.350
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.050
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.270
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.392
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.005
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.050
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.354
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.054
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.315
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.449

I believe this type of AP table is pretty standard, please correct me if I'm wrong. Would someone please be able to show me how to export a reasonably similar thing with darknet? And if the former is not possible, can an AP@0.50:0.95 be exported? Thanks a bunch!

For now I've tried using derivations of the following without success:
./darknet detector map *args -points 101 -iou_thresh 0.50:0.95

[pabsan-0]

Just by chance I found this page. That table can be obtained with the commands shown in there.

[aniket611]

Hi @pabsan-0 , the link seems to be broken. Can you please write here the command to pass other IoU thresholds.
Thanks.

[pabsan-0]

Hey @aniket611 I believe the link pointed to this issue. However, if you're literally after the table shown above you're better off reformatting your results according to the MS COCO standards and then using the official pycocotools. I honestly do not know if it is possible to get the table straight out of darknet, but that was the path I finally took a year ago.