SceneCode: Monocular Dense Semantic Reconstruction Using Learned Encoded Scene Representations.
Shuaifeng ZhiMichael BloeschStefan LeuteneggerAndrew J. DavisonPublished in: CVPR (2019)
Keyphrases
- optical flow fields
- architectural scenes
- image sequences
- dense reconstruction
- shape recovery
- scene flow
- scene reconstruction
- three dimensional
- static scene
- semantic representations
- stereo camera
- urban scenes
- semantic context
- optical flow
- point cloud
- multi view stereo
- depth cues
- relative depth
- d scene
- road vehicles
- intermediate representations
- piecewise planar
- ego motion
- dense sampling
- sift flow
- video sequences
- stereo reconstruction
- pose estimation
- flow field
- image correspondences
- multiple images
- reconstruction method
- scene structure
- structure from motion
- bundle adjustment
- depth map
- multi view
- image motion
- computer vision
- ground plane
- dynamic scenes
- single image
- camera motion
- camera calibration
- real scenes
- affine camera
- scene classification
- stereo correspondence
- calibrated cameras