Data Product Features

Distribution

Format

• MP4 video files
• Annotation files (JSON, CSV, TXT, or equivalent)

Encoding

• H.264 / H.265

Structure

• Organized into folders containing synchronized videos and corresponding annotation files.

Data Volume

• Approximate Duration: 800 hours
• Approximate Storage Size: 5.6 TB (≈5,600 GB)
• Approximate Records: 10,000 annotated recordings

Usage

• Object Detection: Training object detection and localization models.
• Object Tracking: Multi-frame object tracking and motion analysis.
• Scene Understanding: Semantic understanding of indoor commercial and residential environments.
• Action Recognition: Human activity recognition and behavior analysis.
• Robotics: Robot perception and manipulation research.
• Computer Vision: Model training for visual perception tasks.
• Embodied AI: Learning from annotated real-world interactions.
• Academic Research: Benchmarking AI and robotics algorithms.

Coverage

Geographic Coverage

Time Range

• Start Date: Not specified
• End Date: Not specified

Demographics

• Adults (where applicable)
• Mixed gender participants
• Residential environments
• Commercial environments

License

AI Training Rights

• Use the Data Product to train, fine-tune, and evaluate machine learning models, including large language models.
• Incorporate Data Product content into models and commercialize resulting model outputs.
• Create derivative works (model weights, embeddings, feature representations, etc.) for any lawful purpose.

Restrictions

• The Data Product itself may not be sold, redistributed, or shared as a standalone dataset outside of licensed usage.
• Licensee must comply with all applicable laws, including data protection and privacy regulations.

Who Can Use It

• Data Scientists: Train annotated computer vision and AI models.
• Researchers: Conduct research in computer vision, robotics, and machine learning.
• Robotics Engineers: Develop perception, navigation, and manipulation systems.
• Businesses: Build AI-powered vision applications and automation solutions.
• Universities: Support education, benchmarking, and scientific research.

Data Dictionary

Additional Notes

• Collected across residential and commercial environments in Asia.
• Recorded using a GoPro or equivalent camera.
• Videos are available in 4K and 1080p at 30/60 FPS using H.264/H.265 encoding.
• The dataset includes corresponding annotation files for AI and computer vision tasks.
• Suitable for robotics, object detection, tracking, scene understanding, embodied AI, and machine learning research.
• The streaming preview may appear lower quality, but the downloaded files are provided in full-resolution quality.
• Released under the CC0 license for both research and commercial AI applications.