DAB-DETR: Dynamic Anchor Boxes are Better Queries for DETR

Code

• github.com/slongliu/dab-detr
  OfficialIn paperpytorch★ 575
• github.com/IDEA-Research/detrex
  pytorch★ 2,274
• github.com/alibaba/EasyCV
  pytorch★ 1,949
• github.com/idea-research/dn-detr
  pytorch★ 604
• github.com/idea-research/dab-detr
  pytorch★ 575
• github.com/horrible-dong/teamdetr
  pytorch★ 21
• github.com/helq2612/biadt
  pytorch★ 9
• github.com/Tajamul21/Detection-Classification-and-Semantic_Segmentation-of-apples
  pytorch★ 2

Abstract

We present in this paper a novel query formulation using dynamic anchor boxes for DETR (DEtection TRansformer) and offer a deeper understanding of the role of queries in DETR. This new formulation directly uses box coordinates as queries in Transformer decoders and dynamically updates them layer-by-layer. Using box coordinates not only helps using explicit positional priors to improve the query-to-feature similarity and eliminate the slow training convergence issue in DETR, but also allows us to modulate the positional attention map using the box width and height information. Such a design makes it clear that queries in DETR can be implemented as performing soft ROI pooling layer-by-layer in a cascade manner. As a result, it leads to the best performance on MS-COCO benchmark among the DETR-like detection models under the same setting, e.g., AP 45.7\% using ResNet50-DC5 as backbone trained in 50 epochs. We also conducted extensive experiments to confirm our analysis and verify the effectiveness of our methods. Code is available at https://github.com/SlongLiu/DAB-DETR.

Tasks

Benchmark Results

Reproductions