Revisiting DETR for Small Object Detection via Noise-Resilient Query Optimization

TL;DR

This paper introduces a Noise-Resilient Query Optimization framework for small object detection that enhances transformer-based detectors by reducing noise sensitivity and improving query quality, leading to better detection performance.

Contribution

It proposes novel NT-FPN and PS-RPN modules that improve feature fusion and query generation without extra hyperparameters, advancing small object detection methods.

Findings

NRQO outperforms state-of-the-art baselines on multiple benchmarks.

NT-FPN effectively preserves spatial and semantic information during feature fusion.

PS-RPN generates high-quality positive queries with improved anchor-ground truth matching.

Abstract

Despite advancements in Transformer-based detectors for small object detection (SOD), recent studies show that these detectors still face challenges due to inherent noise sensitivity in feature pyramid networks (FPN) and diminished query quality in existing label assignment strategies. In this paper, we propose a novel Noise-Resilient Query Optimization (NRQO) paradigm, which innovatively incorporates the Noise-Tolerance Feature Pyramid Network (NT-FPN) and the Pairwise-Similarity Region Proposal Network (PS-RPN). Specifically, NT-FPN mitigates noise during feature fusion in FPN by preserving spatial and semantic information integrity. Unlike existing label assignment strategies, PS-RPN generates a sufficient number of high-quality positive queries by enhancing anchor-ground truth matching through position and shape similarities, without the need for additional hyperparameters. Extensive experiments on multiple benchmarks consistently demonstrate the superiority of NRQO over state-of-the-art baselines.