An Investigation of Potential Function Designs for Neural CRF.
Zechuan HuYong JiangNguyen BachTao WangZhongqiang HuangFei HuangKewei TuPublished in: EMNLP (Findings) (2020)
Keyphrases
- potential functions
- markov random field
- conditional random fields
- pairwise
- random fields
- maximum entropy
- weighted sums
- network architecture
- superpixels
- neural network
- higher order
- graph cuts
- crf model
- non stationary
- iterative conditional
- energy function
- parameter estimation
- graphical models
- information retrieval
- maximum likelihood
- probabilistic model
- image segmentation
- segmentation method
- image classification
- semi supervised
- multi class
- multiscale
- computer vision