Best Practices for Fine-Tuning Visual Classifiers to New Domains.
Brian ChuVashisht MadhavanOscar BeijbomJudy HoffmanTrevor DarrellPublished in: ECCV Workshops (3) (2016)
Keyphrases
- fine tuning
- viable alternative
- fine tune
- application domains
- support vector
- training set
- visual information
- training samples
- training data
- visual perception
- feature set
- real world
- feature selection
- classification systems
- linear classifiers
- fine tuned
- transfer learning
- test set
- case study
- training examples
- naive bayes
- feature space
- software engineering
- classification algorithm
- high level
- human vision
- supervised classification
- binary classifiers
- sufficient training data
- decision trees
- supervised learning
- visual features
- classification rate
- ensemble learning
- machine learning methods
- classification method
- machine learning algorithms
- svm classifier