fzhan-cloud's repositories
CVPR2023-Paper-Code-Interpretation
cvpr2022/cvpr2021/cvpr2020/cvpr2019/cvpr2018/cvpr2017 论文/代码/解读/直播合集，极市团队整理
hello-world
just another description
ml-visuals
🎨 ML Visuals contains figures and templates which you can reuse and customize to improve your scientific writing.
Retinal-lesion-Segmentation
Diabetic retinopathy (DR) is one of the most important complications of diabetes. Accurate segmentation of DR lesions helps early diagnosis of DR. However, due to the scarcity of pixel-level annotations and the large diversity between different types of DR lesions, the existing deep learning methods are very challenging in performing segmentation on retinal images. In this study, we propose a novel data augmentation method based on Poisson-blending (PB) algorithm to generate synthetic images, which can be easily adapted to other medical anomaly segmentation tasks to alleviate the training data scarcity issue. We also proposed a CNN architecture for the simultaneous segmentation of multiscale anomaly signs. The performances are compared with the state-of-the-art methods on Indian Diabetic Retinopathy Image Dataset (IDRiD) and e-ophtha datasets, both widely used in the research community. The results indicate that the proposed method significantly outperforms the state-of-the-art methods.