AR-NeRF: Unsupervised Learning of Depth and Defocus Effects from Natural Images with Aperture Rendering Neural Radiance Fields.
Takuhiro KanekoPublished in: CoRR (2022)
Keyphrases
- natural images
- global illumination
- unsupervised learning
- depth from defocus
- computer graphics
- object recognition
- shape recovery
- natural scenes
- depth map
- sparse coding
- light field
- defocus blur
- multiscale
- higher order
- image patches
- denoising
- input image
- supervised learning
- primary visual cortex
- receptive fields
- augmented reality
- computer vision
- neural network
- spatial domain
- power spectrum
- image statistics
- natural image statistics
- semi supervised
- depth estimation
- statistics of natural images
- high quality
- deep learning
- active learning
- photorealistic
- reflectance properties
- natural image patches
- blurred images
- depth information
- shape from shading
- high resolution
- point spread function
- real scenes
- deep belief networks
- single image
- machine learning