Convolutional neural network hyperparameters designed for highest object recognition accuracy in satellite imagery.
Povilas GudziusOlga KurasovaErnestas FilatovasPublished in: DATA (2019)
Keyphrases
- recognition accuracy
- satellite imagery
- hyperparameters
- convolutional neural network
- recognition rate
- remote sensing
- face recognition
- change detection
- model selection
- cross validation
- closed form
- satellite images
- bayesian framework
- sample size
- bayesian inference
- support vector
- land cover
- maximum a posteriori
- noise level
- face detection
- maximum likelihood
- random sampling
- em algorithm
- urban areas
- image processing
- error rate
- decision trees
- remote sensing images
- incomplete data
- feature space
- prior knowledge
- prior information
- machine learning