[Feature] Support TOOD: Task-aligned One-stage Object Detection (ICCV 2021 Oral)

[RangiLyu]

TOOD: Task-aligned One-stage Object Detection

https://arxiv.org/abs/2108.07755

Introduction

One-stage object detection is commonly implemented by optimizing two sub-tasks: object classification and localization, using heads with two parallel branches, which might lead to a certain level of spatial misalignment in predictions between the two tasks. In this work, we propose a Task-aligned One-stage Object Detection (TOOD) that explicitly aligns the two tasks in a learning-based manner. First, we design a novel Task-aligned Head (T-Head) which offers a better balance between learning task-interactive and task-specific features, as well as a greater flexibility to learn the alignment via a task-aligned predictor. Second, we propose Task Alignment Learning (TAL) to explicitly pull closer (or even unify) the optimal anchors for the two tasks during training via a designed sample assignment scheme and a task-aligned loss. Extensive experiments are conducted on MS-COCO, where TOOD achieves a 51.1 AP at single-model single-scale testing. This surpasses the recent one-stage detectors by a large margin, such as ATSS (47.7 AP), GFL (48.2 AP), and PAA (49.0 AP), with fewer parameters and FLOPs. Qualitative results also demonstrate the effectiveness of TOOD for better aligning the tasks of object classification and localization.

Checklist

• Implement TOOD.
• Align accuracy.
  tood_r50_fpn_1x_coco:

 Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.424
 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=1000 ] = 0.595
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=1000 ] = 0.461
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.251
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.455
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.555
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.614
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=300 ] = 0.614
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=1000 ] = 0.614
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.416
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.658
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.779

• Add unit tests.
• Upload checkpoints.

[codecov]

Codecov Report

Merging #6746 (0d6c0e2) into dev-v2.20.0 (eb1903f) will increase coverage by 0.33%.
The diff coverage is 85.15%.

@@               Coverage Diff               @@
##           dev-v2.20.0    #6746      +/-   ##
===============================================
+ Coverage        62.04%   62.37%   +0.33%     
===============================================
  Files              322      327       +5     
  Lines            25717    26122     +405     
  Branches          4356     4421      +65     
===============================================
+ Hits             15955    16293     +338     
- Misses            8913     8963      +50     
- Partials           849      866      +17     

Flag	Coverage Δ
unittests	62.35% <85.15%> (+0.33%)	⬆️

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

Impacted Files	Coverage Δ
mmdet/models/losses/focal_loss.py	69.33% <73.07%> (+2.06%)	⬆️
mmdet/models/detectors/tood.py	75.00% <75.00%> (ø)
mmdet/core/hook/set_epoch_info_hook.py	80.00% <80.00%> (ø)
mmdet/models/dense_heads/tood_head.py	83.39% <83.39%> (ø)
mmdet/core/bbox/assigners/task_aligned_assigner.py	93.65% <93.65%> (ø)
mmdet/core/bbox/assigners/__init__.py	100.00% <100.00%> (ø)
mmdet/core/hook/__init__.py	100.00% <100.00%> (ø)
mmdet/models/dense_heads/__init__.py	100.00% <100.00%> (ø)
mmdet/models/detectors/__init__.py	100.00% <100.00%> (ø)
mmdet/models/losses/gfocal_loss.py	97.36% <100.00%> (+0.87%)	⬆️
... and 17 more

---

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 eb1903f...0d6c0e2. Read the comment docs.