IDEA-Research/DAB-DETR
Official PyTorch implementation of DAB-DETR, a transformer-based object detection model published at ICLR 2022.

Velocity · 7d
+0.4
★ / day
Trend
→steady
star history
DAB-DETR introduces dynamic anchor boxes as queries for Detection Transformer (DETR) architectures, improving convergence and accuracy in end-to-end object detection. The implementation provides training and evaluation code for the transformer-based detection model on standard benchmarks. The repository also releases related works including DINO and Mask DINO detection models in their detrex toolbox.