Skip to content
Ivan edited this page Jul 18, 2013 · 1 revision

My function has four inputs and five outputs. The four inputs are: Topics: This will be a cell array of strings. Example: {'hand', 'background'};

Objects: This is a data structure in the form of cells. Each cell in Objects represents one video. Each video has its own number of frames. If you would like to classify images only then each video will have one frame associated with it.

dictionarySize: This designates the number of words in the corpus/code book.

order: My algorithm requires the ground truth. If order is not given, it will display the first image of the video and allow you to select which topic that image belongs to. This will give the ground truth and also return the labels which you gave to the image. You can then plug this order in the next time you run the algorithm and it will automatically assign labels to the objects you want to classify.

There are four outputs associated with my program: missrate: Describes the percent of wrong classifications.

confMatrix: The confusion matrix.

predictedClasses: This gives the best permutation of the labels.

kernel: This is the similarity matrix which describes how similar each video is to another video with respect to objects.

order: This is the order of the labels that you assigned for the videos.

Clone this wiki locally