Improving Reinforcement Learning Based Image Captioning with Natural Language Prior.
Tszhang GuoShiyu ChangMo YuKun BaiPublished in: CoRR (2018)
Keyphrases
- reinforcement learning
- natural language
- input image
- image data
- image representation
- single image
- image classification
- image features
- image analysis
- multiscale
- image content
- image retrieval
- test images
- image collections
- image regions
- segmentation method
- machine learning
- template matching
- prior information
- image pixels
- image segmentation
- prior model
- optimal policy
- vector field
- markov decision processes
- spatial information
- hough transform
- edge detection
- maximum a posteriori
- feature points
- bayesian framework
- natural language processing
- function approximation
- pixel values
- object recognition
- learning algorithm