[Fix]: Alpha of ciou loss should not have gradient.

[RangiLyu]

In the official implementation of CIOU loss, the alpha does not have gradient. And when iou<=0.5, alpha=0. Refer to:

https://github.com/Zzh-tju/CIoU/blob/d8573fc9ef221e89fd0dfeef42027acd764c49ff/layers/modules/multibox_loss.py#L53

[codecov]

Codecov Report

Merging #5835 (b1c65cc) into master (bcd3b50) will decrease coverage by 0.02%.
The diff coverage is 100.00%.

❗ Current head b1c65cc differs from pull request most recent head 1ffc669. Consider uploading reports for the commit 1ffc669 to get more accurate results

@@            Coverage Diff             @@
##           master    #5835      +/-   ##
==========================================
- Coverage   67.09%   67.06%   -0.03%     
==========================================
  Files         301      301              
  Lines       23335    23334       -1     
  Branches     3841     3841              
==========================================
- Hits        15657    15650       -7     
  Misses       6884     6884              
- Partials      794      800       +6     

Flag	Coverage Δ
unittests	67.06% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
mmdet/datasets/pipelines/formating.py	66.37% <100.00%> (ø)
mmdet/models/losses/iou_loss.py	94.76% <100.00%> (+0.05%)	⬆️
mmdet/models/dense_heads/dense_test_mixins.py	41.02% <0.00%> (-2.57%)	⬇️
mmdet/models/roi_heads/test_mixins.py	72.34% <0.00%> (-2.13%)	⬇️
mmdet/datasets/custom.py	63.37% <0.00%> (-0.22%)	⬇️
mmdet/datasets/pipelines/transforms.py	76.47% <0.00%> (-0.14%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 682f03d...1ffc669. Read the comment docs.

[ZwwWayne]

Suggest updating unit test

[ZwwWayne]

Add a faster rcnn model with CIoULoss

[RangiLyu]

Training results of CIOU Faster R-CNN

Before:

Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.378
 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=1000 ] = 0.583
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=1000 ] = 0.408
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.216
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.412
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.490
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.523
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=300 ] = 0.523
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=1000 ] = 0.523
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.324
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.559
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.661 

After:

Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.378
 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=1000 ] = 0.583
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=1000 ] = 0.408
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.216
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.413
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.492
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.521
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=300 ] = 0.521
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=1000 ] = 0.521
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.323
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.561
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.665

Almost the same.

[RangiLyu]

CIOU with weight=12 refering from the paper

 Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.380
 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=1000 ] = 0.583
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=1000 ] = 0.412
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.218
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.418
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.495
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.527
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=300 ] = 0.527
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=1000 ] = 0.527
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.328
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.566
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.670

[ZwwWayne]

can be merged after update config and model link