Diverse Generation while Maintaining Semantic Coordination: A Diffusion-Based Data Augmentation Method for Object Detection

TL;DR

This paper presents a diffusion-based data augmentation method for object detection that improves dataset diversity while maintaining semantic coordination, leading to significant performance gains.

Contribution

It introduces a novel augmentation technique using pre-trained diffusion models, a Category Affinity Matrix, and a Surrounding Region Alignment strategy to balance diversity and semantic consistency.

Findings

Improves object detection AP by +1.4, +0.9, +3.4 on three models

Enhances dataset diversity without losing semantic coordination

Outperforms existing augmentation methods

Abstract

Recent studies emphasize the crucial role of data augmentation in enhancing the performance of object detection models. However,existing methodologies often struggle to effectively harmonize dataset diversity with semantic coordination.To bridge this gap, we introduce an innovative augmentation technique leveraging pre-trained conditional diffusion models to mediate this balance. Our approach encompasses the development of a Category Affinity Matrix, meticulously designed to enhance dataset diversity, and a Surrounding Region Alignment strategy, which ensures the preservation of semantic coordination in the augmented images. Extensive experimental evaluations confirm the efficacy of our method in enriching dataset diversity while seamlessly maintaining semantic coordination. Our method yields substantial average improvements of +1.4AP, +0.9AP, and +3.4AP over existing alternatives on three distinct object detection models, respectively.