LEMON：面向手术场景感知的大规模内窥镜单目数据集与基础模型
LEMON: A Large Endoscopic MONocular Dataset and Foundation Model for Perception in Surgical Settings
LEMON 提供了迄今最大规模的手术视频数据集（938 小时、8500 万帧），并配套一个通过自监督增强知识蒸馏预训练的基础模型 LemonFM；对视觉研究者而言，这是一个可直接用于手术场景下游任务（阶段识别、动作识别、器械检测、语义分割）的大规模预训练资源，其数据聚合流程与自监督蒸馏策略也可迁移到其他标注稀缺的领域。
原文依据 a foundation model pretrained on LEMON using a novel self-supervised augmented knowledge distillation approach
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论传统的开放获取手术数据集通常规模有限，往往包含少于 100 个视频和不足 30 小时的素材，导致模型泛化能力较差。为缓解这一数据限制，本文通过一种新颖的聚合流程从在线来源收集高分辨率视频，构建了名为 LEMON 的新数据集。LEMON 收录了超过 4K 个手术视频，总计 938 小时（8500 万帧）的高质量素材，涵盖多种手术类型，在规模和范围上超越了现有替代方案，并包含两个新颖的下游任务。为证明该多样化数据集的有效性，我们提出了 LemonFM，一个在 LEMON 上通过新颖的自监督增强知识蒸馏方法预训练的基础模型。LemonFM 在四个下游任务和六个数据集上持续优于现有的手术基础模型，在手术阶段识别上取得显著提升（在 AutoLaparo、M2CAI16 和 Cholec80 上的 Jaccard 指标分别提升 +9.5、+9.4 和 +8.4 个百分点），在手术动作识别上提升 +4.4 个百分点（CholecT50 上的平均精度均值），在手术器械存在检测上提升 +5.3 和 +10.2 个百分点（Cholec80 和 GraSP 上的平均精度均值），在手术语义分割上提升 +10.3 个百分点（CholecSeg8k 上的 mDice）。LEMON 和 LemonFM 将成为研究界和工业界的基础资源，加速自主机器人手术系统的研发进展，并最终推动全球更安全、更可及的手术护理。数据集、代码和模型已公开于 https://github.com/visurg-ai/LEMON。
英文原文摘要 1675 字符
Traditional open-access datasets focusing on surgical procedures are often limited by their small size, typically consisting of fewer than 100 videos and less than 30 hours of footage, which leads to poor model generalization. To address this data limitation, a new dataset called LEMON has been compiled using a novel aggregation pipeline that collects high-resolution videos from online sources. Featuring an extensive collection of over 4K surgical videos totaling 938 hours (85 million frames) of high-quality footage across multiple procedure types, LEMON offers a comprehensive resource surpassing existing alternatives in size and scope, including two novel downstream tasks. To demonstrate the effectiveness of this diverse dataset, we introduce LemonFM, a foundation model pretrained on LEMON using a novel self-supervised augmented knowledge distillation approach. LemonFM consistently outperforms existing surgical foundation models across four downstream tasks and six datasets, achieving significant gains in surgical phase recognition (+9.5pp, +9.4pp, and +8.4pp in Jaccard on AutoLaparo, M2CAI16, and Cholec80), surgical action recognition (+4.4pp in mAP on CholecT50), surgical tool presence detection (+5.3pp and +10.2pp in mAP on Cholec80 and GraSP), and surgical semantic segmentation (+10.3pp in mDice on CholecSeg8k). LEMON and LemonFM will serve as foundational resources for the research community and industry, accelerating progress in developing autonomous robotic surgery systems and ultimately contributing to safer and more accessible surgical care worldwide. Dataset, code, and models are publicly available at https://github.com/visurg-ai/LEMON.