On the optical flow model selection through metaheuristics.
Danillo Roberto PereiraJosé DelpianoJoão Paulo PapaPublished in: EURASIP J. Image Video Process. (2015)
Keyphrases
- model selection
- optical flow
- motion segmentation
- cross validation
- hyperparameters
- image sequences
- bayesian learning
- parameter estimation
- regression model
- feature selection
- machine learning
- motion estimation
- motion field
- sample size
- optical flow estimation
- meta learning
- least squares
- flow field
- statistical inference
- selection criterion
- variable selection
- generalization error
- bayesian information criterion
- bayesian methods
- error estimation
- gaussian process
- mixture model
- parameter determination
- motion model
- simulated annealing
- moving objects
- marginal likelihood
- leave one out cross validation
- information criterion
- automatic model selection
- statistical learning
- optimal solution
- computer vision
- factor analysis
- training set
- model selection criteria