Data Editors：Yongxiang Liu

Dataset Introduction: The team from the College of Electronic Science and Technology at the National University of Defense Technology aims to advance the "HUIYAN-ATRNet" Target Sample Dataset to provide researchers with a robust data benchmarking ecosystem. The ATRNet-STAR dataset represents the team's first step toward establishing a large-scale target characteristic database. It represents a breakthrough compared to the previous SAR vehicle target benchmark dataset, MSTAR. The team spent nearly two years completing the project design, data collection and processing, and the construction of methodological benchmarks.

ATRNet-STAR comprises nearly 200,000 target images, featuring 40 target types (covering 4 major vehicle classes, 21 vehicle subclasses, and 40 vehicle types, including most civilian vehicle types such as sedans, SUVs, pickup trucks, buses, trucks, and tankers), diverse scenarios (including urban areas, factories, woodlands, bare soil, and sandstone), various imaging conditions (including different angles, bands, and polarization modes), and multiple formats (including floating-point complex raw data and processed 8-bit magnitude data). The dataset is annotated with detailed information, including target dimensions, target locations, imaging angles, and resolution. It is currently the largest publicly available SAR vehicle recognition dataset, with a scale more than 10 times larger than any previous vehicle dataset. Its abundant target samples can support research in various areas, including generation, detection, and classification.

Additionally, to facilitate research innovation and method comparison, the team has established a meticulously designed classification and detection benchmark, ATRBench, comprising 7 experimental settings (e.g., robust recognition, few-shot recognition, and transfer learning) and 15 representative methods. Experimental results demonstrate that SAR ATR under complex conditions remains highly challenging, while large-scale pre-trained models have shown relatively superior performance. Pre-training on this dataset will improve recognition of different ground targets.

The comprehensive target samples and experimental benchmarks provided by this dataset can serve as a new research platform for SAR ATR, further advancing the field. For further inquiries, please contact Dr. Weijie Li at lwj2150508321@sina.com.

References and Citation Format for This Dataset:

[1] Yongxiang Liu, Weijie Li, Li Liu, Jie Zhou, Bowen Peng, Yafei Song, Xuying Xiong, Wei Yang, Tianpeng Liu, Zhen Liu, Xiang Li. ATRNet-STAR: A Large Dataset and Benchmark Towards Remote Sensing Object Recognition in the Wilde (ATRNet-STAR-1.0) [OL]. Journal of Radars, 2025. https://radars.ac.cn/web/data/getData?newsColumnId=f2d2f517-4d2f-4995-8ca5-b2d8d280f862&pageType=en

[2] Yongxiang Liu, Weijie Li, Li Liu, Jie Zhou, Bowen Peng, Yafei Song, Xuying Xiong, Wei Yang, Tianpeng Liu, Zhen Liu, Xiang Li. ATRNet-STAR: A Large Dataset and Benchmark Towards Remote Sensing Object Recognition in the Wilde[DB/OL]. (2025-03-13)[2025-03-30]. https://arxiv.org/abs/1706.03825.

Arxiv ：https://arxiv.org/abs/2501.13354v4

Data release time：2025-11-13