A benchmark dataset for studying camera-based perception in autonomous driving under varying sensor configurations. It is created using the CARLA simulator to model real-world fleet variations in camera placement, orientation, field of view, and count. The dataset was authored by timb2001 and last updated on 2026-06-25.
Use Cases
- Evaluating domain adaptation methods based on varying camera rig geometries described in the benchmark.
- Training perception models to be invariant to changes in camera placement and orientation mentioned in the description.
- Benchmarking the robustness of vision systems against the cross-rig domain gap introduced by sensor variation.
- Studying the effect of camera count and field-of-view changes on perception performance under controlled conditions.
Strengths
- Focuses on a controlled study of the cross-rig domain gap, a specific and relevant challenge for real-world deployment.
- Uses the CARLA simulator, a standard platform for autonomous driving research, to generate 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.
- Row count and dataset size are unknown, which may limit suitability assessment.
Provenance
- Source
- huggingface user timb2001
- Collection Method
- Simulated using the CARLA autonomous driving simulator.
- Freshness
- Last updated 2026-06-25 14:21:05; freshness should be verified.