Zipf

Public sample of a proprietary multimodal urban driving dataset from ZipF.in.

First-person scooter footage from Indian cities, recorded by participating riders. Day and night. This sample is the horizontal labeled videos only.

Each video is about one minute, 1280×720, with labels drawn on the picture. Faces and license plates are pixelated.

What is labeled

16 labelled classes, including cars, pedestrians, trucks, potholes, and traffic signs.

What this sample does not include

The full package is not in this repository. A licensed copy also includes:

• Labeled frames, with a JSON file for every frame
• Telemetry synced to the video: GPS, speed, acceleration, yaw, and vibration

Non-commercial use of these videos follows CC BY-NC 4.0. For commercial use or the full dataset, request access or email us at founders@zipf.in.