AI-driven rabies diagnosis using transfer learning and data augmentation
A study on arXiv (2604.19823) proposes an automated AI diagnostic system for rabies using fluorescent image analysis. The system employs transfer learning with EfficientNetB0, EfficientNetB2, VGG16, and Vision Transformer (ViTB16) architectures. Three data augmentation strategies were tested on a dataset of 155 microscopic images (123 positive, 32 negative). TrivialAugmentWide proved most effective for model generalization. The research addresses the scarcity of skilled laboratory personnel in African and Asian countries where rabies remains a major public health concern.
Key facts
- arXiv paper 2604.19823
- Automated AI diagnostic system for rabies
- Uses fluorescent image analysis
- Transfer learning with EfficientNetB0, EfficientNetB2, VGG16, ViTB16
- Dataset: 155 microscopic images (123 positive, 32 negative)
- TrivialAugmentWide was most effective augmentation strategy
- Addresses scarcity of skilled personnel in Africa and Asia
- Rabies is a major public health concern in low-data settings
Entities
Institutions
- arXiv