Synthetic Dataset from Semantic Diffusion Models
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/naamiinepal/synthetic-boost
下载链接
链接失效反馈官方服务：
资源简介：
该数据集由语义扩散模型（SDMs）生成的合成图像组成，旨在提升视觉-语言分割模型（VLSMs）在心脏超声分割任务上的表现。该数据集被用于在真实心脏超声图像上进行微调之前，对VLSMs进行预训练，结果显示，这一过程显著提高了性能指标。该任务专注于心脏超声分割。
This dataset consists of synthetic images generated by Semantic Diffusion Models (SDMs), and is designed to improve the performance of Vision-Language Segmentation Models (VLSMs) on the cardiac ultrasound segmentation task. It is used to pre-train VLSMs prior to their fine-tuning on real cardiac ultrasound images, with experimental results showing that this process significantly enhances performance metrics. This task focuses on cardiac ultrasound segmentation.
提供机构：
Semantic Diffusion Models
