All Urban | 31,968 4K Video Clips | 360° POV Driving Footage

Data Product Features

Usage

• Generative video (text-to-video, image-to-video): real, stable 4K camera motion through real streets, with rich metadata for balanced training sets and accurate captions.
• World models and driving simulation: all 9 angles are recorded at the same moment, so models can learn how a scene stays consistent across viewpoints and over time.
• Autonomous driving and ADAS: varied road environments for pre-training and domain generalization.
• Gaussian splats and 3D reconstruction: the nodal 360-degree rig captures overlapping views from 9 synchronized cinema cameras at once. With calibration and approximate route tracks available on request, it works well for Gaussian splats, NeRFs and multi-view 3D reconstruction.

Technical Details

• Rig: 9 synchronized cinema cameras: 8 nodal horizontal cameras at 45-degree intervals plus one overhead camera, shot on Panasonic GH4 (2018–19) and Netflix-approved Panasonic BGH1 cinema cameras (2023 onward) with the Lumix 12-35mm lens at 12 mm, 4096x2160 (C4K), 23.976 fps. Horizontal field of view is about 65° (GH4) and 77.5° (BGH1), so neighboring angles overlap by 20–32°.
• Sync: all 9 angles of every set share matching timecode and in/out points.
• Camera calibration: nominal intrinsics for both camera models are included. Calibrated intrinsics and rig extrinsics are available on request, solved per shoot day for 2023 onward and per season for 2018–19. Lens optical stabilization was on to reduce road vibration.
• Location: every clip is logged with the street traveled, start and end cross streets, and direction. Approximate route tracks (1-second latitude/longitude/heading, GPX or CSV, reconstructed from those logs against OpenStreetMap) are available on request. The footage has no recorded GPS.
• Rights and privacy: self-shot on public roads and 100% owned; nothing scraped or licensed in. Members of the public, vehicles and license plates appear incidentally, and no blurring has been applied. Licensees are responsible for complying with privacy laws in their own jurisdictions.

Coverage

• Geographic Coverage: Los Angeles: Downtown LA, Hollywood, West Hollywood, Hollywood Hills, Laurel Canyon, Koreatown, Westwood & UCLA, Beverly Hills, Bel Air, Culver City, Santa Monica, Malibu, LAX & Westchester, Long Beach, Pasadena, La Cañada Flintridge, Santa Clarita, and the San Fernando Valley (Sherman Oaks, Studio City, Encino, Van Nuys, North Hollywood, Valley Village, Granada Hills, Lake Balboa, Sun Valley). Orange County coast: Newport Beach, Laguna Beach, Dana Point, Huntington Beach, Seal Beach. San Francisco Bay Area: San Francisco, Oakland, Berkeley, Richmond, San Rafael, San Mateo, Burlingame, San Bruno, San Carlos, Redwood City, Palo Alto, Menlo Park, Fremont, Hayward, San Leandro, Union City. San Diego. New York City: Manhattan, Brooklyn, Queens and the Bronx. Washington DC. Chicago. Atlanta & Alpharetta. Austin. Las Vegas. New Orleans. Denver. Memphis. Portland. Seattle. Albuquerque. Cheyenne. Minneapolis. Columbia, SC. Richmond, VA. Canada: Calgary and Winnipeg. Smaller California cities: Stockton, San Luis Obispo, Santa Maria.
• Time Range: July 2018 - November 2024.

License

• 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 the resulting models and model outputs.
• Create derivative works (model weights, embeddings, etc.) for any lawful purpose.

• The Data Product itself (the raw footage and metadata) may not be sold, sublicensed, redistributed or shared outside the licensee's organization, or used as stock or production footage.
• Licensee must comply with all applicable laws, including data protection and privacy regulations.

Who Can Use It

• AI labs and video-generation teams: for training and evaluating generative video and world models.
• Autonomous driving and robotics teams: for perception pre-training and testing in new road environments.
• 3D and mapping teams: for Gaussian splats, NeRFs, digital twins and visual place recognition.
• Researchers: for academic or scientific studies (academic licenses available on request).

Data Dictionary

• OriginalFileName: DP230622-024-Chicago_Illinois-FC.MOV
• SetID: DP230622-024 (shoot date + shot number; all 9 angles of a set share it)
• Camera: angle code and direction (FC 0°, FR 45°, SR 90°, BR 135°, BC 180°, BL 225°, SL 270°, FL 315°, UP overhead)
• StartTimecode: matching across all 9 angles of a set
• Description: text description
• Base: Driving plate, driving, plate, POV, Point of view
• Date: 6-digit date in format YYMMDD (230622)
• Num: shot number for that date (24)
• Time: day, night, dawn, dusk (day)
• Eve: morning, afternoon, evening (afternoon)
• Season: Spring, Summer, Autumn, Winter (summer)
• Weather: Sunny, Cloudy, Overcast (sunny, Clear)
• Region: Illinois
• City: Chicago
• Urban: City, Urban, Downtown
• Setting: Desert, Brush, Marshland, Lake, River, Ocean
• Travel On: Michigan Avenue
• From: West Ida B. Wells Drive
• To: East Monroe Street
• Direction: North, Northbound
• Road: Street
• Traffic: heavy, congested, stop and go, static, stop
• Features: Buildings, Art Institute of Chicago
• Notes: Bug, Dirty
• Friendly Title: Front driving plate - Michigan Avenue Northbound at Chicago