Active Self-Training for Weakly Supervised 3D Scene Semantic Segmentation.
Gengxin LiuOliver van KaickHui HuangRuizhen HuPublished in: CoRR (2022)
Keyphrases
- d scene
- weakly supervised
- semantic segmentation
- semi supervised
- superpixels
- object class
- topic models
- single image
- depth map
- relation extraction
- optical flow
- conditional random fields
- long range
- multiscale
- computer vision
- named entities
- high quality
- segmentation method
- action recognition
- probabilistic model
- viewpoint
- three dimensional
- data sets