SceneCode: Monocular Dense Semantic Reconstruction using Learned Encoded Scene Representations.
Shuaifeng ZhiMichael BloeschStefan LeuteneggerAndrew J. DavisonPublished in: CoRR (2019)
Keyphrases
- optical flow fields
- architectural scenes
- image sequences
- optical flow
- three dimensional
- scene flow
- stereo camera
- scene reconstruction
- d scene
- point cloud
- dense reconstruction
- static scene
- urban scenes
- intermediate representations
- semantic representations
- flow field
- depth cues
- sift flow
- dense sampling
- shape recovery
- stereo reconstruction
- relative depth
- piecewise planar
- single image
- uncalibrated cameras
- ego motion
- wide baseline
- road vehicles
- dense correspondence
- video sequences
- image correspondences
- apparent motion
- pose estimation
- image reconstruction
- motion field
- scene structure
- projective geometry
- multi view
- dynamic scenes
- affine camera
- stereo correspondence
- semantic context
- depth map
- camera calibration
- calibrated cameras
- real scenes
- depth information
- motion estimation