Switching to Discriminative Image Captioning by Relieving a Bottleneck of Reinforcement Learning.
Ukyo HondaTaro WatanabeYuji MatsumotoPublished in: CoRR (2022)
Keyphrases
- image data
- reinforcement learning
- image content
- single image
- input image
- multiscale
- image retrieval
- image features
- image analysis
- template matching
- feature points
- image segmentation
- low level
- image classification
- image representation
- hough transform
- spatial information
- image regions
- vector field
- region of interest
- image sequences
- image structure
- pixel values
- learning algorithm
- pictorial structure
- natural images
- edge detection
- state space
- high resolution
- multi agent
- similarity measure
- feature extraction
- feature selection