SynFundus-1M: A High-quality Million-scale Synthetic Fundus Images Dataset with Fifteen Types of Annotation
Core Concepts
SynFundus-1M is a high-quality synthetic fundus image dataset with sophisticated annotations, beneficial for medical imaging research.
Abstract
SynFundus-1M released by Baidu Inc. contains over one million synthetic fundus images with 15 types of annotation.
The dataset aims to enhance early detection and treatment of eye diseases through deep learning methods.
Extensive experiments demonstrate the authenticity and effectiveness of SynFundus-1M in disease diagnosis models.
The dataset offers a large scale of images and comprehensive annotations compared to existing datasets.
Models trained on SynFundus-1M show superior performance and faster convergence in downstream tasks.
"Extensive experiments prove that our synthetic images can hardly be distinguished from the authentic ones even by four experienced annotators."
"We hope that this study will fuel further breakthroughs and broaden the application scope of the fundus imaging analysis."