VL-SAT: Visual-Linguistic Semantics Assisted Training for 3D Semantic Scene Graph Prediction in Point Cloud.
Ziqin WangBowen ChengLichen ZhaoDong XuYang TangLu ShengPublished in: CVPR (2023)
Keyphrases
- point cloud
- real world objects
- laser scanner
- semantic representations
- stereo camera
- urban scenes
- semantic information
- high level semantics
- natural language
- structure from motion
- architectural scenes
- surface reconstruction
- dominant plane
- point sets
- multi view stereo
- dense reconstruction
- image sequences
- three dimensional
- d scene
- point cloud data
- training set
- video sequences
- cad model
- single image
- visual features
- training samples
- viewpoint
- logic programming
- outdoor scenes