rhoban/model_learning_old

Rhoban Model Learning

Immediate TODO

• Implement ArucoCube (map of faces) and ArucoCubeCollection
  • And add test for tagsSheet, arucoCube etc...
• Implement Priors class and CompositePriors
• Implement shortcut names (at least "all" and for compositeModels)

Taxonomy

• Sample: A couple composed of an Input and an Observation
• DataSet: Two sets of samples, one for training and one for validation
• DataSetReader: An entity allowing to read the data inside a file to extract a DataSet
• Model: Parametrized entities, model learning is finding the 'best' set of parameters
• Predictors: for a given model and a given input
• Prior: A distribution of probability on the parameters of a model

Targeted Models

• Mocap
• PosesOptimization
• PosesInference

Untested learning model

• Calibration Model (compiles but not tested)

Disabled learnings (due to new model)