翻译：在 ＤＲ分类中对于质量较差的原始图像没有进行剔除导致在数据实时扩增阶段产生较多的无用特征数据影响模型的效率和精度后续的任务就是研究一套完整的视网膜图像质量评估方法将着重注意提高数据集的质量。同时探讨研究更适合图像差异小的样本如视网膜图像的注意力机制进一步提高模型的分类性能。

In the DR classification, the poor-quality original images were not eliminated, resulting in a large amount of useless feature data during the real-time data augmentation phase, which affects the efficiency and accuracy of the model. The subsequent task is to develop a complete method for evaluating the quality of retinal images, with a focus on improving the quality of the dataset. At the same time, attention mechanisms that are more suitable for samples with small image differences (such as retinal images) will be explored to further improve the classification performance of the model