Deep Octree-based CNNs with Output-Guided Skip Connections for 3D Shape and Scene Completion.
Peng-Shuai WangYang LiuXin TongPublished in: CoRR (2020)
Keyphrases
- single image
- d objects
- three dimensional
- d scene
- video sequences
- multiple images
- quadtree
- cellular neural networks
- scene understanding
- data structure
- dynamic scenes
- image sequences
- surface geometry
- complex scenes
- multi view
- input image
- neural network
- visual hull
- cast shadows
- viewpoint
- wire frame
- geometric primitives
- real scenes
- stereo reconstruction
- image measurements
- cad model
- scene structure
- man made
- object models
- object detection
- spatial relations
- high resolution
- camera motion
- depth map
- input data