MECCANO

MECCANO数据集是由卡塔尼亚大学的IPLAB实验室与Next Vision s.r.l.合作创建的，专注于工业类似环境中的人类行为理解。该数据集包含20个多模态以自我为中心的视频，涵盖了注视信号、深度图和RGB视频，这些数据是通过定制的头戴设备在意大利和英国同时采集的。数据集特别标注了用于识别和预测人与物体交互的基本任务，适用于研究如动作识别、主动物体检测与识别、以自我为中心的人-物体交互检测、动作预测和下一个主动物体检测等任务。此外，数据集还提供了关于动作和交互的时间注释，以及空间边界框注释，以支持更深入的人类行为分析。

The MECCANO dataset was co-developed by the IPLAB Laboratory at the University of Catania and Next Vision s.r.l., focusing on human behavior understanding in industrial-like environments. This dataset contains 20 multimodal egocentric video samples covering gaze signals, depth maps and RGB videos, which were collected simultaneously in Italy and the United Kingdom using custom-built head-mounted devices. The dataset specifically provides annotations for fundamental tasks related to identifying and predicting human-object interactions, and is applicable to research work including action recognition, active object detection and recognition, egocentric human-object interaction detection, action prediction and next active object detection. In addition, the dataset also offers temporal annotations for actions and interactions, as well as spatial bounding box annotations to support more in-depth human behavior analysis.