Conceptual 12M: Pushing Web-Scale Image-Text Pre-Training To Recognize Long-Tail Visual Concepts.
Soravit ChangpinyoPiyush SharmaNan DingRadu SoricutPublished in: CVPR (2021)
Keyphrases
- visual concepts
- web scale
- web images
- image content
- image collections
- image annotation
- long tail
- image search
- image data
- image features
- image retrieval
- input image
- image classification
- semantic gap
- image segmentation
- test images
- visual features
- supervised learning
- keypoints
- video content
- object categories
- visual content
- visual data
- low level
- information retrieval
- semantic information
- keywords
- training set
- high resolution
- positive examples
- data sources
- image database
- training examples