Annotated Indian Road Footage for Autonomous Driving

Indian Road Driving Dataset is the largest open dataset of annotated Indian road footage, created by ThirdEye Labs. It addresses a critical gap by capturing unique challenges like dense mixed traffic, auto-rickshaws, and cattle, which are absent from datasets like BDD100K and nuScenes.

Use Cases

• Train vehicle detection models on annotated footage featuring auto-rickshaws and cattle to improve object detection in mixed traffic.
• Develop semantic segmentation models for road scenes with informal lane usage and extreme lighting conditions.
• Benchmark autonomous driving perception systems against dense traffic scenarios specific to Indian roads.
• Analyze driver behavior and traffic patterns from dashcam footage to model unpredictable road interactions.

Strengths

• Largest open dataset of annotated Indian road footage, addressing a known data gap.
• Captures unique road features like auto-rickshaws, cattle, and informal lane usage not found in other major datasets.
• Designed for computer vision tasks including object detection and image segmentation.

Limitations

• Specific scale metrics like row count, file size, and annotation counts are unknown.
• Geographic coverage is limited to Indian roads, which may limit generalizability to other regions.
• Dataset freshness and update frequency are unclear beyond the last platform update date.

Provenance

Source

ThirdEye Labs.

Collection Method

Annotated road footage from dashcams, likely for autonomous driving research.

Freshness

Last platform update was 2026-03-23.

Geography

India.