Wild-Places: Large-Scale LiDAR Point Clouds for Natural Place Recognition

Wild-Places provides large-scale LiDAR point cloud data for place recognition in unstructured natural environments, released by CSIRO Robotics in 2023. The dataset serves as a benchmark for loop-closure and localization algorithms in settings lacking the geometric regularity of urban areas.

Use Cases

• Evaluating loop-closure detection using LiDAR point clouds in environments without man-made landmarks
• Training deep learning models for place recognition using unstructured natural terrain features
• Benchmarking SLAM algorithms on LiDAR sequences collected in forest and trail settings

Strengths

• Peer-reviewed methodology published in IEEE ICRA 2023
• Focuses on unstructured natural environments, a domain underrepresented in LiDAR research
• Large-scale data volume specifically curated for place recognition benchmarks

Limitations

• Geographic bias likely limited to the specific natural regions where CSIRO conducted data collection
• Lack of structured geometric features like buildings or signs limits utility for urban-specific research

Provenance

Source

CSIRO Robotics / IEEE ICRA 2023

Collection Method

LiDAR sensor

Freshness

Last updated in March 2026; based on 2023 ICRA publication research.