MM-OR-SG-Dataset
收藏MM-OR-SG-Dataset 数据集概述
数据集描述
- 来源:基于MultiModal Operating Room Dataset for Semantic Understanding of High-intensity Surgical Environments的精选子集。
- 用途:支持两个任务的小型项目:
- 场景图生成
- 事件重要性分类
- 新增内容:事件重要性标签("routine", "noteworthy", "critical")为伪标签,通过SVC生成,不属于原始MM-OR数据集。
数据结构
- 特征:
images:图像序列,每个样本包含四张图像(三张连续手术视频帧 + 一张全景分割掩码,尺寸均为256×256)。annotations:JSON字符串,包含:rel_annotations:场景图三元组列表(主体, 关系, 客体)。event_significance:事件重要性标签("routine", "noteworthy", "critical")。
- 数据划分:
train:867个样本,大小约313.18 MB。
使用示例
python from datasets import load_dataset
ds = load_dataset("fiqqy/MM-OR-SG-Dataset")
sample = ds["train"][0]
images = sample["images"]
import json
annotations = json.loads(sample["annotations"])
print("Scene Graph Relations:", annotations["rel_annotations"])
print("Event Significance:", annotations["event_significance"])
引用信息
bibtex @inproceedings{ozsoy2025mm, title={Mm-or: A large multimodal operating room dataset for semantic understanding of high-intensity surgical environments}, author={{"O}zsoy, Ege and Pellegrini, Chantal and Czempiel, Tobias and Tristram, Felix and Yuan, Kun and Bani-Harouni, David and Eck, Ulrich and Busam, Benjamin and Keicher, Matthias and Navab, Nassir}, booktitle={Proceedings of the Computer Vision and Pattern Recognition Conference}, pages={19378--19389}, year={2025} }




