Scholarlink

Diabetic retinopathy (DR) is a side effect to diabetes that damaging the retina and can lead to blindness. Deep learning (DL) has been suggested as an automated method for detecting diabetic retinopathy (DR) due to its powerful models. In this study, DenseNet121 and EfficientNetB0 Models were used with the weights of the previous training on the ImageNet as initialization, utilizing APTOS 2019 dataset (3662 color retinal images) that was preprocessed through Gaussian filters and resized to (224,224). Then during the training, data augmentation and resampling were implemented. These models obtained (92 and 95%) testing accuracies for DenseNet121 and EfficientNetB0, respectively.

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

Automatic Diabetic Retinopathy Identification Based on Deep Learning Models

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Fatima Jabbar Jwad,
  • Ashwan A. Abdulmunem

摘要

Diabetic retinopathy (DR) is a side effect to diabetes that damaging the retina and can lead to blindness. Deep learning (DL) has been suggested as an automated method for detecting diabetic retinopathy (DR) due to its powerful models. In this study, DenseNet121 and EfficientNetB0 Models were used with the weights of the previous training on the ImageNet as initialization, utilizing APTOS 2019 dataset (3662 color retinal images) that was preprocessed through Gaussian filters and resized to (224,224). Then during the training, data augmentation and resampling were implemented. These models obtained (92 and 95%) testing accuracies for DenseNet121 and EfficientNetB0, respectively.