STRIDE-QA: Visual Question Answering for Autonomous Driving in Tokyo

Tokyo driving data provides a large-scale visual question answering dataset for physically grounded spatiotemporal reasoning. It contains 16 million question-answer pairs over 270,000 frames, constructed from 100 hours of multi-sensor driving data. The dataset was created by turing-motors and last updated on the platform in January 2026.

Use Cases

• Training VQA models for autonomous vehicles based on multi-sensor driving data.
• Developing spatiotemporal reasoning algorithms based on dense annotations like 3D bounding boxes.
• Benchmarking multimodal AI systems on physically grounded driving scenarios.
• Analyzing object interactions and tracks in urban driving environments.

Strengths

• Large scale with 16 million QA pairs.
• Dense annotations include 3D bounding boxes, segmentation masks, and multi-object tracks.
• Based on 100 hours of real-world multi-sensor driving data.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• The mini version is a preliminary release and may not match the full dataset's format.

Provenance

Source

turing-motors

Collection Method

Constructed from multi-sensor driving data.

Freshness

Last updated 2026-01-23 08:10:58; freshness should be verified.

Geography

Tokyo