MDS-DETR: DETR with Masked Duplicate Suppressor
Researchers have developed MDS-DETR, a novel object detection model that improves upon the DEtection TRansformer (DETR) architecture. MDS-DETR addresses DETR's slow convergence and low recall issues by integrating both one-to-one and one-to-many label assignment within a single decoder. This is achieved through a Masked Duplicate Suppressor (MDS) that filters redundant predictions, leading to more efficient and accurate object detection. AI
IMPACT MDS-DETR offers improved training efficiency and accuracy for object detection tasks, potentially benefiting applications in computer vision.