Scholarlink

One of the main causes of visual loss in diabetic individuals worldwide is diabetic retinopathy. In order to stop irreparable visual impairment, early detection and prompt management are essential. Deep learning algorithms have proven uplifting results in automating the detection of DR from retinal pictures in recent years. Using the IDRiD dataset, this study examines the efficacy of the EfficientNetB1 model for diabetic retinopathy (DR) identification. Diabetic patients frequently experience vision impairment due to DR, which highlights the importance of precise and effective detection techniques. The EfficientNetB1 model is refined on the IDRiD dataset by utilizing transfer learning, which allows it to pick up discriminative features linked to different DR stages. The EfficientNetB1 model outperforms the ResNet50 and InceptionResNetV2 architectures in terms of DR detection, according to a comparative investigation. These results highlight the promise of utilizing cutting-edge deep learning architectures for accurate diagnosis of DR, allowing for early intervention to reduce the negative effects of diabetes on vision health.

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

Diabetic Retinopathy Detection Using Deep Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • G. Sucharitha,
  • G. Sreeja,
  • P. Anil,
  • Sachi Nandan Mohanty,
  • S. K. Danish

摘要

One of the main causes of visual loss in diabetic individuals worldwide is diabetic retinopathy. In order to stop irreparable visual impairment, early detection and prompt management are essential. Deep learning algorithms have proven uplifting results in automating the detection of DR from retinal pictures in recent years. Using the IDRiD dataset, this study examines the efficacy of the EfficientNetB1 model for diabetic retinopathy (DR) identification. Diabetic patients frequently experience vision impairment due to DR, which highlights the importance of precise and effective detection techniques. The EfficientNetB1 model is refined on the IDRiD dataset by utilizing transfer learning, which allows it to pick up discriminative features linked to different DR stages. The EfficientNetB1 model outperforms the ResNet50 and InceptionResNetV2 architectures in terms of DR detection, according to a comparative investigation. These results highlight the promise of utilizing cutting-edge deep learning architectures for accurate diagnosis of DR, allowing for early intervention to reduce the negative effects of diabetes on vision health.