TIMo-A Dataset for Indoor Building Monitoring with a Time-of-Flight Camera.

Area of Science:

• Computer Vision
• Robotics
• Human-Computer Interaction

Background:

• Existing surveillance video datasets primarily offer grayscale or RGB data.
• Depth information, crucial for 3D scene understanding, is underrepresented in current video datasets.
• Time-of-flight (ToF) cameras provide depth data, valuable for detailed spatial analysis.

Purpose of the Study:

• To introduce TIMo, a new dataset for indoor monitoring using time-of-flight (ToF) camera data.
• To address the scarcity of depth video datasets for surveillance applications.
• To facilitate research in person detection, people counting, and anomaly detection.

Main Methods:

• Collected depth videos of individuals performing predefined actions using a ToF camera.
• Recorded data across two distinct indoor locations with varied camera perspectives (top-down and tilted).
• Provided detailed annotations for actions and individuals within the depth videos.

Main Results:

• The TIMo dataset contains depth videos with annotated human actions.
• Experimental evaluation results from baseline algorithms are included.
• The dataset captures indoor environments from different viewpoints.

Conclusions:

• TIMo dataset fills a critical gap for depth-based indoor monitoring research.
• The dataset enables advancements in computer vision tasks like person detection and anomaly detection.
• TIMo facilitates the development of more robust and accurate indoor surveillance systems.