Welcome!
3D Vision, Large scale, Deep Learning, Point Cloud, Segmentation, MLS & ALS
Get startedWHAT CAN YOU DO
Semantic Segmentation
The points are labelled according to the category they belong to. For example, all the points of different cars have the same label.
Instance Segmentation
The points are labelled according to objects. For example, points of a different cars has different labels even though they are both cars.
Object Detection
For countable classes, every object has one 3D minimal bounding box, which could help to differentiate different points of different objects.
ABOUT This Dataset
WHU-Urban3D
Driven by a strong desire to bridge the gap between the increasing demand for 3D outdoor scene understanding and the limited datasets and approaches for LiDAR (e.g., MLS and ALS) point cloud instance segmentation, we introduce a manually annotated large-scale 3D dataset, named WHU-Urban3D, for semantic and instance segmentation. WHU-Urban3D has the following features which makes it distinct from other existing datasets: 1) contains both MLS and ALS point cloud; 2) covers large-scale road and urban scene; 3) includes rich categories in real world; 4) involving point-wise instance and semantic labels. WHU-Urban3D covers 3.6 × 106 m2 and contains over 300 million points, which will be continuously enlarged. For more detail, you could refer to docs or online preview (Do NOT use Safari for online viewing!!!).
