Figure 6
From: A generalised framework for detailed classification of swimming paths inside the Morris Water Maze
Workflow diagram illustrating the analysis procedure. Cyan boxes indicate automatic process; orange boxes indicate objects of importance or results; Partial Labelling box (magenta) implies extensive user interaction with the process; the grey box groups the processes taking part in the classification procedure. After the trajectory segmentation, eight trajectory features for each segment are computed and a certain number of segments are manually labelled. Afterwards a pool of classifiers is generated. ‘Strong’ classifiers (cross validation error <25%) are then selected from the pool and work together (majority voting) as a team (ensemble) to produce the classification results. Throughout the process the labelling quality is constantly assessed and in case of weak classification results we go back to the labelling stage.
