About this document
Diabetic Retinopathy Classification with ViT by ziadsalahmohamed is a document available to read on EtoBox.
This research article presents a new model using Vision Transformer (ViT) with Masked Autoencoders (MAE) for classifying referable diabetic retinopathy (DR) from large-size retinal images. The study demonstrates that pre-training ViT on over 100,000 retinal images significantly improves classification performance compared to models pre-trained on ImageNet, achieving an accuracy of 93.42% and an area under the curve (AUC) of 0.9853. This approach reduces the number of required images for effective training w
- Author
- ziadsalahmohamed
- Language
- EN