Confident Pseudo-labeled Diffusion Augmentation for Canine Cardiomegaly Detection
收藏资源简介:
该数据集由叶史瓦大学的研究团队创建,旨在通过合成数据增强和伪标签技术解决犬类心脏肥大检测中的训练数据不足问题。数据集包含3000张合成的犬类胸部X光图像,每张图像均手动标注了脊椎心脏评分(VHS)。数据生成过程利用先进的扩散模型生成高质量的合成图像,并通过伪标签策略进一步优化数据集。该数据集的应用领域为兽医诊断,特别是犬类心脏肥大的自动化检测,旨在提高诊断的准确性和效率,减少人工标注的工作量。
This dataset was developed by a research team at Yeshiva University to address the shortage of training data for canine cardiac hypertrophy detection using synthetic data augmentation and pseudo-labeling techniques. It comprises 3,000 synthetic canine chest X-ray images, each manually annotated with the Vertebral Heart Score (VHS). The data generation pipeline utilizes state-of-the-art diffusion models to generate high-quality synthetic images, and further optimizes the dataset via pseudo-labeling strategies. Targeted at veterinary diagnostic applications, particularly automated detection of canine cardiac hypertrophy, this dataset aims to enhance diagnostic accuracy and efficiency while reducing the burden of manual annotation work.




