Fine-grained Image Captioning with CLIP Reward.
Jaemin ChoSeunghyun YoonAjinkya KaleFranck DernoncourtTrung BuiMohit BansalPublished in: CoRR (2022)
Keyphrases
- fine grained
- coarse grained
- image data
- input image
- multiscale
- single image
- image retrieval
- image content
- image features
- edge detection
- tightly coupled
- massively parallel
- access control
- image classification
- low level
- image representation
- image regions
- image matching
- reinforcement learning
- image sequences
- image segmentation
- graphical models
- databases
- high resolution
- relational databases
- energy function
- information retrieval