Tunnels Dataset, Factory Dataset, Corridors Dataset

本研究贡献了三个基于图像的数据集，用于室内机器人探索场景中的异常检测。这些数据集包含超过20万标记帧，覆盖多种异常类型，如灰尘、湿顶和细根等。数据集由模拟和真实世界场景中的无人机和地面机器人采集，分为训练和测试集，用于训练基于自编码器的异常检测模型。这些数据集旨在帮助机器人识别未知环境中的潜在危险，从而提高自主操作的安全性。

This study contributes three image-based datasets for anomaly detection in indoor robotic exploration scenarios. These datasets contain over 200,000 labeled frames and cover a variety of anomaly types, such as dust, wet ceilings, fine roots, and others. Collected by unmanned aerial vehicles (UAVs) and ground robots in both simulated and real-world scenarios, the datasets are split into training and test sets for training autoencoder-based anomaly detection models. These datasets aim to help robots identify potential hazards in unknown environments, thereby improving the safety of autonomous operations.