Rethinking the Misalignment Problem in Dense Object Detection

TL;DR

This paper introduces SALT, a novel operator for dense object detection that spatially disentangles and aligns features for classification and localization tasks, improving detection accuracy.

Contribution

The paper proposes SALT, a plug-in operator that spatially disentangles task-specific features and a self-distillation loss, leading to significant improvements in dense object detection performance.

Findings

SALT improves AP by approximately 2 points across various detectors.

SALT-Net achieves 53.8 AP on MS-COCO test-dev with Res2Net-101-DCN backbone.

The methods enhance the alignment between classification and localization features.

Abstract

Object detection aims to localize and classify the objects in a given image, and these two tasks are sensitive to different object regions. Therefore, some locations predict high-quality bounding boxes but low classification scores, and some locations are quite the opposite. A misalignment exists between the two tasks, and their features are spatially entangled. In order to solve the misalignment problem, we propose a plug-in Spatial-disentangled and Task-aligned operator (SALT). By predicting two task-aware point sets that are located in each task's sensitive regions, SALT can reassign features from those regions and align them to the corresponding anchor point. Therefore, features for the two tasks are spatially aligned and disentangled. To minimize the difference between the two regression stages, we propose a Self-distillation regression (SDR) loss that can transfer knowledge from the refined regression results to the coarse regression results. On the basis of SALT and SDR loss, we propose SALT-Net, which explicitly exploits task-aligned point-set features for accurate detection results. Extensive experiments on the MS-COCO dataset show that our proposed methods can consistently boost different state-of-the-art dense detectors by $\sim$2 AP. Notably, SALT-Net with Res2Net-101-DCN backbone achieves 53.8 AP on the MS-COCO test-dev.