SIEVE: Multimodal Dataset Pruning Using Image Captioning Models.
Anas MahmoudMostafa ElhoushiAmro AbbasYu YangNewsha ArdalaniHugh LeatherAri MorcosPublished in: CoRR (2023)
Keyphrases
- image content
- input image
- image data
- single image
- image features
- image classification
- image dataset
- template matching
- random fields
- image structure
- high resolution
- parametric models
- image representation
- low level
- image analysis
- multiscale
- segmentation method
- image matching
- statistical model
- bayesian framework
- image retrieval
- scale spaces
- image collections
- image pixels
- image regions
- image database
- image segmentation
- pruning method
- object models
- visual effects
- million images
- street view
- computer vision
- weakly labeled
- pixel values
- region of interest
- hough transform
- visual features
- medical images
- edge detection
- digital images
- probabilistic model
- image processing