Run on hardware
- Include RandomForestClassifier with float
- Use embedded model in train/test in notebook
- Run performance test on device
- Implement microphone input on device
- Implement sending of classification from device
- Implement sending of features,audio for subsets of samples
Improve performance in non-matched conditions
- Estimate non-matched conditions by using folds by dataset
- Implement spherical K-means feature learning. Ref Stowell LifeCLEF 2014.
MLCluster
- Setup without LoadBalancer. Probably using NodePort
--set service.type=NodePort - Create script for easily getting right addresses for
NodePort