Task-wise Sampling Convolutions for Arbitrary-Oriented Object Detection in Aerial Images

TL;DR

This paper introduces task-wise sampling convolutions (TS-Conv), a novel AOOD method that adaptively samples features for localization and classification, improving detection accuracy and robustness in aerial images.

Contribution

The paper proposes TS-Conv with task-specific sampling and a dynamic label assignment strategy, enhancing orientation robustness and detection performance in AOOD tasks.

Findings

TS-Conv outperforms existing methods on multiple datasets

Improved orientation robustness of features

Demonstrated scalability across various scenes and categories

Abstract

Arbitrary-oriented object detection (AOOD) has been widely applied to locate and classify objects with diverse orientations in remote sensing images. However, the inconsistent features for the localization and classification tasks in AOOD models may lead to ambiguity and low-quality object predictions, which constrains the detection performance. In this article, an AOOD method called task-wise sampling convolutions (TS-Conv) is proposed. TS-Conv adaptively samples task-wise features from respective sensitive regions and maps these features together in alignment to guide a dynamic label assignment for better predictions. Specifically, sampling positions of the localization convolution in TS-Conv are supervised by the oriented bounding box (OBB) prediction associated with spatial coordinates, while sampling positions and convolutional kernel of the classification convolution are designed to be adaptively adjusted according to different orientations for improving the orientation robustness of features. Furthermore, a dynamic task-consistent-aware label assignment (DTLA) strategy is developed to select optimal candidate positions and assign labels dynamically according to ranked task-aware scores obtained from TS-Conv. Extensive experiments on several public datasets covering multiple scenes, multimodal images, and multiple categories of objects demonstrate the effectiveness, scalability, and superior performance of the proposed TS-Conv.