VIsCUIT: Visual Auditor for Bias in CNN Image Classifier.
Seongmin LeeJudy HoffmanZijie J. WangDuen Horng ChauPublished in: CVPR (2022)
Keyphrases
- low level
- image data
- visual perception
- input image
- single image
- image representation
- multiscale
- image content
- visual appearance
- image features
- image retrieval
- image analysis
- semantic labels
- visual effects
- image segmentation
- image collections
- visual attributes
- human observers
- web images
- visual cues
- region of interest
- test images
- image regions
- segmentation method
- high resolution
- feature set
- keypoints
- image set
- neural network
- training data
- visual features
- visually similar
- human visual
- image classification
- learning algorithm
- feature extractor
- pixel classification
- visual concepts
- feature points
- natural images
- extracted features
- pixel values
- segmentation algorithm
- spatial relations