Coverage and Quality Driven Training of Generative Image Models.
Konstantin ShmelkovThomas LucasKarteek AlahariCordelia SchmidJakob VerbeekPublished in: CoRR (2019)
Keyphrases
- image data
- single image
- image analysis
- image content
- input image
- image representation
- image retrieval
- multiscale
- random fields
- image features
- set of training images
- edge detection
- discriminative classifiers
- image pixels
- neural network
- image collections
- bayesian framework
- image regions
- segmentation method
- data driven
- probabilistic model
- low level
- high resolution
- high quality
- similarity measure
- spatial information
- image statistics
- prior knowledge
- quality assessment
- object detectors
- labeled images
- information redundancy
- image quality assessment
- visual effects
- structured prediction
- low quality
- pixel values
- test images
- segmentation algorithm
- generative model
- training data
- image sequences
- image segmentation