Finger vein biometric identification is gaining prominence due to its reliability and security. However, challenges persist, particularly in handling low-quality images, that inspires for improvement despite advancements in state-of-the-art methods. Recent state-of-the-art methods can perform better still there is a gap to achieve more. In this paper, we investigate use of transfer learning with well-known CNN models to analyse their performance in person identification using finger vein biometric. Comprehensive experiments were conducted using the FV-USM and THU-FVFDT2 datasets, which consist of person finger vein images captured under controlled scenarios. The datasets were split into training, validation, and testing sets using a 70–20-10 ratio. CNN models, including VGG19, MobileNetV2, InceptionV3, and EfficientNetB0, were assessed based on classification metrics such as accuracy, precision, recall, F1-score, and training time. Among these, VGG19 achieved an impressive 99.9% and 99.67% accuracy for FV-USM and THU-FVFDT2 datasets respectively. The experimental results demonstrate the effectiveness of these models in person identification and their superiority over traditional shallow feature-based finger vein biometric approaches.