Rethinking Transformer-based Set Prediction for Object Detection

Abstract

DETR is a recently proposed Transformer-based method which views object detection as a set prediction problem and achieves state-of-the-art performance but demands extra-long training time to converge. In this paper, we investigate the causes of the optimization difficulty in the training of DETR. Our examinations reveal several factors contributing to the slow convergence of DETR, primarily the issues with the Hungarian loss and the Transformer cross-attention mechanism. To overcome these issues we propose two solutions, namely, TSP-FCOS (Transformer-based Set Prediction with FCOS) and TSP-RCNN (Transformer-based Set Prediction with RCNN). Experimental results show that the proposed methods not only converge much faster than the original DETR, but also significantly outperform DETR and other baselines in terms of detection accuracy.

arXiv commentAccepted to ICCV 2021

Cited by in corpus (5)

• Masked-attention Mask Transformer for Universal Image Segmentation
• Exploiting Both Domain-specific and Invariant Knowledge via a Win-win Transformer for Unsupervised Domain Adaptation
• Contrastive Learning for Local and Global Learning MRI Reconstruction
• Armour: Generalizable Compact Self-Attention for Vision Transformers
• End-to-End Entity Detection with Proposer and Regressor