IMAGINATOR: Pre-Trained Image+Text Joint Embeddings using Word-Level Grounding of Images.
Varuna KrishnaS. SuryavardanShreyash MishraSathyanarayanan RamamoorthyParth PatwaMegha ChakrabortyAman ChadhaAmitava DasAmit P. ShethPublished in: CoRR (2023)
Keyphrases
- pre trained
- input image
- word level
- image features
- image retrieval
- image classification
- training data
- image analysis
- image matching
- image set
- training examples
- handwritten documents
- lighting conditions
- information retrieval
- language independent
- illumination conditions
- feature points
- document analysis
- image segmentation
- machine translation
- text retrieval
- target object
- sentence level
- control signals
- image content
- statistical model
- single image
- visual features
- text mining
- small number
- relevance feedback
- active learning