MEDISEG
收藏资源简介:
MEDISEG是由伦敦大学圣乔治学院与帝国理工学院联合构建的药品图像分割数据集,旨在解决现实场景中药物识别因光照、遮挡和多药混杂导致的误差问题。该数据集包含8,262张图像,涵盖32种药片类型,提供精细的实例分割标注,覆盖单药片至多药片混杂(最多13片/帧)的复杂场景。数据通过iPhone 12 Pro Max采集,模拟真实用药环境中的光照变化和剂量盒遮挡,并采用COCO格式标注。其创新性在于首次系统性地捕捉现实用药场景的视觉复杂性,为开发抗干扰的AI药物识别系统提供关键训练资源,特别适用于老年多药治疗等高风险场景的用药安全研究。
MEDISEG is a pharmaceutical image segmentation dataset jointly constructed by St George's, University of London and Imperial College London, aiming to address the recognition errors caused by illumination variations, occlusion and multi-drug mixing in real-world medication scenarios. This dataset contains 8,262 images covering 32 types of tablets, with fine-grained instance segmentation annotations, and covers complex scenarios ranging from single-tablet cases to mixed multi-tablet scenes (up to 13 tablets per frame). The data was collected using an iPhone 12 Pro Max, simulating illumination changes and blister pack occlusion in real medication environments, and all annotations are in COCO format. Its innovation lies in the first systematic capture of the visual complexity of real medication scenarios, providing a critical training resource for developing anti-interference AI drug recognition systems, and it is particularly suitable for medication safety research in high-risk scenarios such as polypharmacy treatment for the elderly.




