Improving mAP

[nirbenz]

Heya @xuzheyuan624 -continuing our discussion from the other repository - I have extended your code to support varying aspect ratios and letterboxing. Doing this I was able to get an mAP of:

DONE (t=5.06s). Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.324 Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.572 Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.331 Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.190 Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.352 Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.421 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.277 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.414 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.432 Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.276 Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.458 Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.547

Which is almost the same as the original Darknet implementation of YOLOv3 - and equivalent to ultralytics's implementation.

I switched to testing your code due to what seems to be working training code + multi GPU support. I will update if/when I get something converged (I've got 9 1080ti's for this purpose).

[xuzheyuan624]

@nirbenz I did the same test when u told me to keep aspect ratio, (I haven't update the code to github yet), but I got mAP(IoU=0.5)=0.565, which is lower than yours.So could u share the code?

[nirbenz]

I actually wrapped your code with an abstraction of my own, which is the same one I used for other detectors I'm testing. I need to make a standalone version for sharing so it'll take me a few days.

Did you have any progress w.r.t. training?

[houweidong]

@xuzheyuan624 look your name， very like chinese， can you give me a qq number， it's my very honor to commute with dashen. - . -