Vision Transformers and Hybrid Models Lead in Retinal Disease Screening Benchmarks

Researchers have benchmarked various deep learning architectures, including convolutional neural networks, vision transformers, hybrid models, and vision-language models, for multi-disease retinal screening. The study utilized the Retinal Fundus Multi-disease Image Dataset (RFMiD) and evaluated performance on binary screening and multi-label classification tasks. Attention-based models, specifically SwinTiny and hybrid architectures like CoAtNet0 and MaxViTTiny, demonstrated superior performance in both binary and multi-label settings. Vision-language models were competitive but did not outperform the best transformer and hybrid backbones. The findings offer a reproducible reference for selecting models for automated retinal screening tools intended for clinical deployment. AI

IMPACT Hybrid and transformer models show promise for improving automated retinal disease screening accuracy.

RANK_REASON The cluster contains an academic paper detailing research findings and benchmarks. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• CLIP ViT-B/16
• CoAtNet0
• MaxViTTiny
• Messidor-2
• Retinal Fundus Multi-disease Image Dataset (RFMiD)
• SigLIP-Base384
• SwinTiny

• paper
• product
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →