MultiCorrupt: A Multi-Modal Robustness Dataset for LiDAR-Camera Fusion

MultiCorrupt is a benchmark dataset for evaluating the robustness of multi-modal 3D object detection models in autonomous driving. The dataset was created by researchers from RWTH Aachen University and TU Berlin and was last updated on March 27, 2025. It focuses on testing LiDAR-camera fusion models against various corruptions.

Use Cases

• Benchmarking model robustness to sensor corruptions based on the dataset's focus on LiDAR-camera fusion corruptions.
• Evaluating 3D object detection performance under adverse conditions based on the dataset's design for robustness testing.
• Developing and testing sensor fusion algorithms based on the multi-modal (LiDAR and camera) nature of the data.

Strengths

• Specifically designed as a robustness benchmark for a critical autonomous driving task.
• Created by academic institutions (RWTH Aachen University and TU Berlin) with expertise in the field.
• Focuses on a modern, multi-modal sensor fusion approach combining LiDAR and camera data.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and sample data are unknown, which may limit suitability assessment.
• The license is unspecified, requiring verification before use.

Provenance

Source

Institute for Automotive Engineering, RWTH Aachen University, and Department of Electrical Engineering and Computer Science, TU Berlin.

Collection Method

Likely created for benchmarking purposes, as indicated by the title and abstract.

Time Range

null

Freshness

Last updated 2025-03 27 10:08:59.

Geography

null