Decoupled Box Proposal and Featurization with Ultrafine-Grained Semantic Labels Improve Image Captioning and Visual Question Answering.
Soravit ChangpinyoBo PangPiyush SharmaRadu SoricutPublished in: EMNLP/IJCNLP (1) (2019)
Keyphrases
- question answering
- semantic labels
- image content
- web images
- image data
- low level
- information retrieval
- image features
- multiscale
- information extraction
- natural language processing
- image retrieval
- visual content
- natural language
- image classification
- image representation
- low level features
- image collections
- named entities
- high level
- machine learning
- news video
- image regions
- feature extraction
- multi modal
- visual information
- spatial relations
- image search
- visual features
- knowledge base
- text classification