REBORN: Reinforcement-Learned Boundary Segmentation with Iterative Training for Unsupervised ASR.
Liang-Hsuan TsengEn-Pei HuDavid Cheng-Han ChiangYuan TsengHung-yi LeeLin-shan LeeShao-Hua SunPublished in: CoRR (2024)
Keyphrases
- learning algorithm
- supervised learning
- reinforcement learning
- iterative learning
- boundary extraction
- training data
- segmentation algorithm
- image segmentation
- segmentation result
- active learning
- unsupervised learning
- unsupervised manner
- boundary estimation
- segmentation method
- supervised training
- region growing
- text segmentation
- hand tuned
- deep architectures
- set of training images
- level set
- multiscale
- bayesian image segmentation
- region based segmentation
- boundary information
- supervised methods
- segmentation of natural images
- semi supervised
- training process
- training set
- texture segmentation
- energy function
- training phase
- object segmentation
- segmentation of textured images
- shape prior
- object boundaries
- speech recognition
- prior shape
- image regions
- training stage
- learning stage
- segmented images
- noisy environments
- segmented regions