AGVBench: A Reliability-Oriented Benchmark of Data Augmentation for Vein Recognition
English summary
AGVBench is a comprehensive benchmark evaluating 30 data augmentation strategies on five public palm- and finger-vein datasets with seven backbone architectures, including CNNs, vision transformers, and vein-specific models. Multi-image mixing methods such as MixUp, PuzzleMix, and StarMixup achieve the highest recognition accuracy but exhibit poor calibration and high vulnerability to adversarial perturbations. Severe geometric transformations often degrade performance, likely due to feature misalignment or spatial cropping. The results demonstrate that accuracy-centric evaluation is insufficient for biometric data augmentation, emphasizing the need for security and robustness. AGVBench provides standardized protocols and open-source code to advance reproducible and secure vein recognition research.
Chinese summary
AGVBench是一个综合性基准，在五个公开的掌静脉和指静脉数据集上，使用七种骨干架构（包括CNN、视觉Transformer和静脉专用模型）评估了30种数据增强策略。多图像混合方法（MixUp、PuzzleMix、StarMixup）取得了最高的识别精度，但校准性差且容易受到对抗扰动攻击。剧烈的几何变换通常因特征错位或空间裁剪导致性能下降。结果表明，仅以精度为中心的评价不足以满足生物特征数据增强的需求，必须考虑安全性和鲁棒性。AGVBench提供了标准化协议和开源代码，以推动可复现且安全的静脉识别研究。
Key points
Multi-image mixing augmentations (MixUp, PuzzleMix, StarMixup) yield top recognition accuracy but suffer from poor calibration and adversarial vulnerability, exposing a performance-security trade-off.
多图像混合增强（MixUp、PuzzleMix、StarMixup）带来了最高的识别精度，但校准性差且易受对抗攻击，凸显了性能与安全之间的权衡。
Severe geometric transformations frequently harm vein recognition performance, likely due to feature misalignment and spatial cropping.
剧烈的几何变换通常会损害静脉识别性能，可能源于特征错位和空间裁剪。
The benchmark demonstrates that accuracy alone is insufficient for evaluating augmentation in biometrics; standardized protocols and code are released for reliable, reproducible research.
该基准表明，仅靠精度不足以评价生物特征增强；提供了标准化协议和代码，以实现可靠且可复现的研究。