A novel weight initialization with adaptive hyper-parameters for deep semantic segmentation.
Nuhman ul HaqAhmad KhanZia ur RehmanAhmad DinLing ShaoSajid ShahPublished in: Multim. Tools Appl. (2021)
Keyphrases
- semantic segmentation
- hyperparameters
- model selection
- cross validation
- closed form
- conditional random fields
- superpixels
- em algorithm
- scene classification
- weakly supervised
- bayesian inference
- random sampling
- maximum likelihood
- bayesian framework
- maximum a posteriori
- support vector
- noise level
- object class
- prior information
- object classes
- incremental learning
- object categories
- image processing
- sample size
- multiscale
- topic models
- expectation maximization
- k means