AnemiaVision uses smartphone photos for non-invasive anemia detection

Researchers have developed AnemiaVision, a web-based system capable of detecting anemia using smartphone images of the palpebral conjunctiva and fingernail beds. The system fine-tunes an EfficientNet-B3 model, incorporating advanced augmentation techniques like TrivialAugmentWide and Mixup, alongside a cosine-annealing learning schedule. AnemiaVision achieved a validation accuracy of 96.2% and an AUC-ROC of 0.98, demonstrating its potential as a low-cost, accessible screening tool for healthcare workers in resource-limited areas. The source code and the system itself are publicly available. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Provides a low-cost, accessible AI-powered screening tool for anemia, potentially improving global health outcomes.

RANK_REASON Academic paper detailing a new AI-driven diagnostic tool.

Read on arXiv cs.CV →

• AnemiaVision
• EfficientNet-B3
• TrivialAugmentWide
• Mixup
• Flask
• PostgreSQL
• Render
• arXiv

• paper
• product
• other