Align-DETR: Enhancing End-to-end Object Detection with Aligned Loss

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Cai, Zhi, Liu, Songtao, Wang, Guodong, Ge, Zheng, Zhang, Xiangyu, Huang, Di
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912166740754432
author Cai, Zhi
Liu, Songtao
Wang, Guodong
Ge, Zheng
Zhang, Xiangyu
Huang, Di
author_facet Cai, Zhi
Liu, Songtao
Wang, Guodong
Ge, Zheng
Zhang, Xiangyu
Huang, Di
contents DETR has set up a simple end-to-end pipeline for object detection by formulating this task as a set prediction problem, showing promising potential. Despite its notable advancements, this paper identifies two key forms of misalignment within the model: classification-regression misalignment and cross-layer target misalignment. Both issues impede DETR's convergence and degrade its overall performance. To tackle both issues simultaneously, we introduce a novel loss function, termed as Align Loss, designed to resolve the discrepancy between the two tasks. Align Loss guides the optimization of DETR through a joint quality metric, strengthening the connection between classification and regression. Furthermore, it incorporates an exponential down-weighting term to facilitate a smooth transition from positive to negative samples. Align-DETR also employs many-to-one matching for supervision of intermediate layers, akin to the design of H-DETR, which enhances robustness against instability. We conducted extensive experiments, yielding highly competitive results. Notably, our method achieves a 49.3% (+0.6) AP on the H-DETR baseline with the ResNet-50 backbone. It also sets a new state-of-the-art performance, reaching 50.5% AP in the 1x setting and 51.7% AP in the 2x setting, surpassing several strong competitors. Our code is available at https://github.com/FelixCaae/AlignDETR.
format Preprint
id arxiv_https___arxiv_org_abs_2304_07527
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Align-DETR: Enhancing End-to-end Object Detection with Aligned Loss
Cai, Zhi
Liu, Songtao
Wang, Guodong
Ge, Zheng
Zhang, Xiangyu
Huang, Di
Computer Vision and Pattern Recognition
DETR has set up a simple end-to-end pipeline for object detection by formulating this task as a set prediction problem, showing promising potential. Despite its notable advancements, this paper identifies two key forms of misalignment within the model: classification-regression misalignment and cross-layer target misalignment. Both issues impede DETR's convergence and degrade its overall performance. To tackle both issues simultaneously, we introduce a novel loss function, termed as Align Loss, designed to resolve the discrepancy between the two tasks. Align Loss guides the optimization of DETR through a joint quality metric, strengthening the connection between classification and regression. Furthermore, it incorporates an exponential down-weighting term to facilitate a smooth transition from positive to negative samples. Align-DETR also employs many-to-one matching for supervision of intermediate layers, akin to the design of H-DETR, which enhances robustness against instability. We conducted extensive experiments, yielding highly competitive results. Notably, our method achieves a 49.3% (+0.6) AP on the H-DETR baseline with the ResNet-50 backbone. It also sets a new state-of-the-art performance, reaching 50.5% AP in the 1x setting and 51.7% AP in the 2x setting, surpassing several strong competitors. Our code is available at https://github.com/FelixCaae/AlignDETR.
title Align-DETR: Enhancing End-to-end Object Detection with Aligned Loss
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2304.07527