Kyushu University Releases MPO Outdoor 3D Datasets Hitting 96% Place Recognition
Two open 3D LiDAR datasets from Kyushu University tackle outdoor place categorization with dense static scans and sparse driving scans across six categories.
ProTakeaways
- Two open outdoor 3D point cloud datasets for place categorization across six semantic categories.
- Dense MPO: 650 FARO Focus3D scans at 9M points with synchronized color, reflectance and range panoramas.
- Sparse MPO: 34,200 Velodyne HDL-32E scans captured while driving, with spherical camera and GPS data.
- LBP on reflectance reaches 96.42 percent on Dense; majority vote pushes Sparse to 89.67 percent.
- Reflectance panoramas outperform color panoramas, a strong signal for laser intensity as a feature.
- Available via Kyushu University robotics lab, complementing KITTI for outdoor semantic scene work.
Teaching a mobile robot to answer the question of where it is turns out to be harder outdoors than in. Indoor RGB-D benchmarks like NYU-Depth and SUN RGB-D have driven steady progress on 3D scene classification, but outside, sensor range collapses and the datasets thin out. A team from Kyushu University is trying to close that gap with two panoramic 3D point cloud datasets aimed squarely at outdoor place categorization.
The Multi-modal Panoramic 3D Outdoor (MPO) datasets come in two flavors, one static and dense, one dynamic and sparse. Both cover the same six semantic categories: forest, coast, indoor parking lot, outdoor parking lot, residential area and urban area, collected around Fukuoka, Japan.

Why outdoor 3D is a data desert
Indoor 3D benchmarks lean heavily on RGB-D cameras because they are cheap and produce dense depth frames, but the sensors have a short detectable range, which is why 3D place datasets for outdoor environments have been scarce. KITTI is the obvious outdoor reference point, but it was built for SLAM, object detection and tracking. Its category labels (city, residential, road, campus) were never designed for semantic place classification.
The authors give a practical reason to prefer LiDAR outside: time-of-flight laser scanners hold up under bright sunlight, darkness, and sudden illumination changes. That matters for autonomous vehicles and outdoor robots that have to work at noon on a beach and inside a dim parking garage on the same route.