MECCANO
收藏资源简介:
MECCANO数据集是首个专注于工业类似环境中第一人称视角下人类-物体交互的研究数据集。该数据集由20名参与者在模拟工业场景中构建摩托车玩具模型的视频组成,包含299,376帧视频数据。数据集不仅标注了时间上的动作片段,还标注了空间上的活跃物体边界框,涵盖了12种动词、20种名词和61种独特动作的分类。MECCANO数据集旨在推动工业环境中第一人称视角下人类动作识别、活跃物体检测、活跃物体识别及第一人称视角下人类-物体交互检测等任务的研究。
The MECCANO dataset is the first research dataset focused on human-object interactions from a first-person perspective in industrial-like environments. It consists of videos of 20 participants assembling a toy motorcycle model in a simulated industrial scenario, totaling 299,376 video frames. The dataset not only annotates temporal action segments but also spatial bounding boxes of active objects, covering 12 verb categories, 20 noun categories and 61 unique action categories. The MECCANO dataset aims to advance research on tasks such as first-person human action recognition, active object detection, active object recognition and first-person human-object interaction detection in industrial environments.




