Data Augmentation Method Based on Context-Aware Patch for Remote Sensing Object Detection

Yogendra Rao Musunuri

doi:10.1109/lgrs.2026.3729279

Data Augmentation Method Based on Context-Aware Patch for Remote Sensing Object Detection

2026-09-02

0

PRE

AI

Y

Yogendra Rao Musunuri

Sun-Yuan Kung (Sun‐Yuan Kung)

DOI：10.1109/lgrs.2026.3729279

Original

Original request for help

Save

Abstract

En 中文

Object detection in aerial imagery has garnered significant attention due to its crucial role in applications such as urban planning, environmental monitoring, and disaster response. However, class imbalance remains a persistent challenge, as minority categories are represented by fewer instances than dominant classes in most remote-sensing datasets. To address this issue, we propose a novel context-aware patch (CAP)-based data augmentation method that mitigates class imbalance by selectively choosing donor images and semantically compatible object instances. The proposed method features two key innovations. First, it preserves contextual coherence by leveraging class-specific donor collections and enforcing geometric and semantic constraints (e.g., boundary, shape, and class). Second, it introduces a class-balancing threshold computed from dataset statistics to dynamically regulate augmentation copy–paste probability for minority categories. We evaluate the approach on DOTA-v1.0, demonstrating a significant improvement in mAP@0.5 and mAP@0.5–0.95 for minority categories while preserving competitive performance for majority classes.

Keywords:

Aerial imagery

class-balancing threshold

class-imbalance

context-aware

object detection

remote sensing

Data Augmentation Method Based on Context-Aware Patch for Remote Sensing Object Detection

Abstract

Journal

Organization

Cited Papers