EPINET: A Fully-Convolutional Neural Network Using Epipolar Geometry for Depth From Light Field Images.
Changha ShinHae-Gon JeonYoungjin YoonIn So KweonSeon Joo KimPublished in: CVPR (2018)
Keyphrases
- epipolar geometry
- feature points
- light field
- geometric constraints
- affine epipolar geometry
- scene depth
- convolutional neural network
- input image
- scene geometry
- image correspondences
- image based rendering
- fundamental matrix
- three dimensional
- epipolar lines
- image retrieval
- point correspondences
- image pairs
- camera motion
- light field rendering
- object recognition
- image features
- defocus blur
- camera positions
- epipolar constraint
- multiple images
- structure from motion
- image registration
- depth information
- d scene
- machine learning
- image sequences
- multi view
- depth map
- image formation
- image matching