Non-biased background dataset, Moving camera dataset

本研究引入了两个新的无偏见背景数据集：一个通过精心组合的相机镜头获得，另一个通过移动机器人采集。这些数据集旨在解决传统计算机视觉数据集中背景偏见和相机运动效应的问题。数据集包含36个视频样本，分别记录了11种不同的人类活动，适用于评估模型在不同背景和相机运动条件下的性能。创建过程中，数据集通过移动机器人和精心设计的背景采集，确保了数据的无偏性和多样性。这些数据集主要用于机器人视觉系统中的异常行为检测，以提高工作场所的安全性。

This study introduces two novel unbiased background datasets: one acquired via carefully configured camera setups, and the other collected using a mobile robot. These datasets are designed to address the issues of background bias and camera motion artifacts present in traditional computer vision datasets. The datasets consist of 36 video samples that record 11 distinct human activities, and are suitable for evaluating model performance under varying background and camera motion conditions. During their development, the datasets were collected via mobile robots and meticulously designed backgrounds to guarantee their unbiasedness and diversity. These datasets are primarily utilized for abnormal behavior detection in robotic vision systems, aiming to enhance workplace safety.