DriveLM: Driving with Graph Visual Question Answering Data

DriveLM-Data comprises two distinct components: DriveLM-nuScenes and DriveLM-CARLA. The dataset is designed to facilitate Perception, Prediction, Planning, Behavior, and Motion tasks with human-written reasoning logic. It was created by OpenDriveLab and was last updated on March 4, 2025.

Use Cases

• Train models for Graph Visual Question Answering (GVQA) based on the described graph-style structure connecting QA pairs.
• Develop autonomous driving agents that perform Perception, Prediction, and Planning tasks based on the multi-task framework.
• Benchmark models that integrate human-written reasoning logic for driving behavior and motion planning.
• Research the connection between visual scene understanding and sequential decision-making in simulated and real-world driving environments.

Strengths

• Dataset is built upon the prevailing nuScenes dataset, suggesting a foundation in a known autonomous driving benchmark.
• Explicitly supports five core autonomous driving tasks (Perception, Prediction, Planning, Behavior, Motion).
• Last updated on 2025-03-04, indicating recent maintenance.

Limitations

• Row count, file formats, and dataset size are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• License information is unknown, requiring verification before use.

Provenance

Source

OpenDriveLab via Hugging Face.

Collection Method

Constructed based on the nuScenes dataset and CARLA simulator.

Freshness

Last updated 2025-03-04 17:15:24.