Scholarlink

The efficient management and prevention of complete vision loss are based upon the prompt recognition of ocular diseases. The present research compared three deep learning neural networks: InceptionV3, ResNet101, and EfficientNetB0 for detection and multiclass classification of five ocular diseases: Age-related Macular Degeneration (AMD), Myopia (M), Eperetinal Membrane (ERM), Retinal Vein Occlusion(RVO) and Drusen(D) from fundus images.The Multilabeled Ocular Disease Intelligent Recognition (ODIR5K) dataset and RfMid Dataset were used for trainig the models. In order to enhance the accuracy of classification, transfer learning techniques were implemented to optimize each model. The results indicate that EficientNetB0 achieved the highest accuracy of 91%, followed by InceptionV3 at 90% and & ResNet101 at 76%. The findings suggest that these models are capable of accurately identifying ocular maladies, thereby offering an effective tool for early treatment and prevention.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multiclass Classification of Retinal Diseases Using Transfer Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Zameer Fatima,
  • Parneeta Dhaliwal,
  • Deepak Gupta

摘要

The efficient management and prevention of complete vision loss are based upon the prompt recognition of ocular diseases. The present research compared three deep learning neural networks: InceptionV3, ResNet101, and EfficientNetB0 for detection and multiclass classification of five ocular diseases: Age-related Macular Degeneration (AMD), Myopia (M), Eperetinal Membrane (ERM), Retinal Vein Occlusion(RVO) and Drusen(D) from fundus images.The Multilabeled Ocular Disease Intelligent Recognition (ODIR5K) dataset and RfMid Dataset were used for trainig the models. In order to enhance the accuracy of classification, transfer learning techniques were implemented to optimize each model. The results indicate that EficientNetB0 achieved the highest accuracy of 91%, followed by InceptionV3 at 90% and & ResNet101 at 76%. The findings suggest that these models are capable of accurately identifying ocular maladies, thereby offering an effective tool for early treatment and prevention.