@Article{cmc.2026.087314, AUTHOR = {Jichao Xie, Xinlei Liu, Tong Duan, Baolin Li, Zhen Zhang, Peng Yi}, TITLE = {3DGe-Aug: 3D Generative Model-Based Data Augmentation for Few-Shot Object Detection}, JOURNAL = {Computers, Materials \& Continua}, VOLUME = {}, YEAR = {}, NUMBER = {}, PAGES = {{pages}}, URL = {http://www.techscience.com/cmc/online/detail/28153}, ISSN = {1546-2226}, ABSTRACT = {In few-shot object detection scenarios involving emerging, rare, or specialized objects, the lack of diverse training samples severely constrains model performance. Recently, data augmentation methods leveraging 2D text-to-image generative models have gained traction; however, they exhibit limited capabilities in generating fine-grained categories and under-represented objects. To address these limitations, we propose 3DGe-Aug, a 3D generative model-based data augmentation framework for few-shot object detection. Our framework constructs 3D assets from limited 2D images via a 3D generative pipeline, thereby overcoming the rendering limitations of 2D generative models for fine-grained objects. Building upon these 3D assets, multi-perspective object images with rich viewpoints and high spatial-geometric consistency are rendered through precise camera pose control. Furthermore, we design an instance pasting algorithm that incorporates multi-scale transformations and dynamic occlusion elimination mechanisms to synthesize training data. Comprehensive experiments conducted across two representative object detection scenarios—unmanned aerial vehicles (UAVs) and autonomous driving—validate the effectiveness of the proposed method across various detection architectures, such as RT-DETR and the YOLO series. To facilitate future research, we also contribute a few-shot object detection benchmark dataset comprising three base classes and six novel classes, along with a corresponding 3D asset set. Our code and datasets are publicly available at https://github.com/xiejichao/DataAug.}, DOI = {10.32604/cmc.2026.087314} }