3D Can Be Explored In 2D : Pseudo-Label Generation for LiDAR Point Clouds Using Sensor-Intensity-Based 2D Semantic Segmentation.
Andrew CaunesThierry ChateauVincent FrémontPublished in: IV (2024)
Keyphrases
- point cloud
- semantic segmentation
- laser scanner
- conditional random fields
- superpixels
- structure from motion
- scene classification
- lidar data
- point cloud data
- weakly supervised
- point sets
- object categories
- image set
- multi label
- image registration
- long range
- object classes
- object recognition
- vision system
- markov random field
- hidden markov models