Chapter 1: Introduction

• 1.1. This manual
• 1.2. Introduction to perClass
• 1.2.1. Versions
• 1.2.2. System requirements
• 1.2.3. Useful general commands
• 1.2.3.1. Displaying perClass version and license information
• 1.2.3.2. Demo examples
• 1.2.3.3. Provide direct feedback to PR Sys Design
• 1.2.3.4. Control messages displayed by perClass
• 1.3. Release notes

• 2.1. Installing free perClass Lite
• 2.2. Installing commercial/academic personal version (hardware dongle)
• 2.3. Installing commercial/academic group version (floating license)
• 2.3.1. Starting the license server
• 2.3.2. Using perClass on client machines

• 3.1. Loading data
• 3.2. Training a fruit classifier
• 3.3. Executing the classifier on new data
• 3.4. Performing crisp decisions
• 3.5. Building a fruit detector
• 3.6. Creating a detector/classifier cascade
• 3.7. Executing the classifier in application outside of Matlab

• 4.1. Introduction
• 4.2. Basic handling of labels
• 4.3. Creating labels
• 4.3.1. Providing per-sample class names
• 4.3.2. Consecutive labeling
• 4.3.3. One entry per class
• 4.3.4. Using label list
• 4.4. Operations on labels
• 4.4.1. Accessing label information
• 4.4.2. Retrieving class sizes and priors
• 4.4.3. Searching for samples with specific labels
• 4.4.4. Subsets of labels by index
• 4.4.5. Relabeling: Changing class names and defining meta-classes
• 4.4.6. Concatenating label sets
• 4.5. Creating label lists
• 4.5.1. From list of class names
• 4.5.2. Using string prefix
• 4.6. Operations on label lists
• 4.6.1. Accessing list content
• 4.6.2. Converting between names and indices

• 5.1. Introduction
• 5.2. Constructing data sets
• 5.3. Importing and exporting data sets
• 5.3.1. Importing data sets
• 5.3.1.1. Importing multiple properties
• 5.3.2. Exporting data sets
• 5.4. Basic operations of data sets
• 5.4.1. Accessing samples, feature, and classes
• 5.4.2. Accessing raw data
• 5.4.3. Accessing class labels
• 5.5. Data set properties
• 5.5.1. Displaying available properties
• 5.5.2. Retrieving properties
• 5.5.3. Setting properties
• 5.5.3.1. Sample properties
• 5.5.3.2. Feature properties
• 5.5.3.3. Data properties
• 5.5.4. Testing presence of a property
• 5.5.5. Removing properties
• 5.6. Multiple sets of labels
• 5.7. Selecting data subsets by property values
• 5.8. Selecting data subsets randomly
• 5.8.1. Splitting data for training and testing
• 5.8.2. Fixing the random state
• 5.9. Renaming classes and defining meta-classes
• 5.9.1. Adding prefix or suffix to all class names

• 6.1. Interactive scatter plot
• 6.1.1. Legend
• 6.1.2. Changing features
• 6.1.3. Sample inspector
• 6.1.4. Switching between different sets of labels
• 6.1.5. Visualizing subsets of samples
• 6.1.6. Bringing class to top, z-order of classes
• 6.1.7. Hand-painting class labels
• 6.1.8. Renaming classes
• 6.1.9. Visualizing live feature distributions in scatter plot
• 6.2. Interactive plot of per-class feature distributions
• 6.3. Working with image data
• 6.3.1. Visualizing images with sdimage
• 6.3.2. Creating image data sets objects directly
• 6.3.3. Hand-painting class labels
• 6.3.4. Cropping images
• 6.3.5. Saving image data set to workspace
• 6.3.6. Working with image subsets
• 6.3.7. Creating image matrix from a data set
• 6.3.8. Storing multiple images in data sets
• 6.3.9. Connecting sdimage and sdscatter
• 6.3.10. Clustering image with k-means
• 6.3.11. Defining connected components

• 7.1. Introduction
• 7.2. Working with image grid
• 7.2.1. Computing features on image grid
• 7.2.2. Visualization of feature images computed on a grid
• 7.3. Types of local image features
• 7.3.1. Local mean and standard deviation
• 7.3.2. Local histograms
• 7.3.3. Features computed from local histograms
• 7.3.4. Co-occurrence matrices
• 7.3.5. User-defined feature extractors
• 7.4. Propagating image labels
• 7.5. Computing features for arbitrary image regions
• 7.5.1. Defining a mask matrix
• 7.5.2. Computing features on a data set subset
• 7.6. Creating label images

• 8.1. Introduction
• 8.1.1. Training a statistical model
• 8.2. Working with pipelines
• 8.2.1. Pipeline output types
• 8.2.2. Pipeline labels
• 8.2.3. Pipeline list
• 8.2.4. Executing the trained model on new data
• 8.2.5. Visualizing model outputs using sdscatter
• 8.3. Pipeline objects
• 8.3.1. Training a pipeline
• 8.3.2. Inspecting pipeline structure and parameters
• 8.3.3. Constructing pipelines manually

• 9.1. Introduction
• 9.2. Default operating point
• 9.2.1. Visualization of decisions in multi dimensional feature space
• 9.3. Defining operating points
• 9.4. Using operating points
• 9.4.1. Example on weighting-based decision
• 9.4.2. Example on thresholding-based decision
• 9.5. Setting specific operating points
• 9.6. Confusion matrices
• 9.6.1. Normalized confusion matrices
• 9.6.2. Storing confusion matrices as strings
• 9.6.3. Rectangular confusion matrices
• 9.6.4. Confusion matrices for a set of operating points
• 9.6.5. Visualization of the per class errors

• 10.1. Introduction
• 10.2. Using sdroc objects
• 10.2.1. Setting the current operating point
• 10.2.2. Performing decisions based on ROC
• 10.2.3. Interactive visualization of ROC decisions
• 10.2.4. Interactive visualization of confusion matrices
• 10.2.5. Accessing operating points
• 10.2.6. Accessing the estimated performances
• 10.2.7. Using different performance measures
• 10.3. Multi-class ROC Analysis
• 10.4. ROC Analysis using target thresholding (detection)
• 10.5. Selecting application-specific operating point
• 10.5.1. Applying performance constraints
• 10.5.2. Constraints using the low-level methods
• 10.5.3. Cost-sensitive optimization
• 10.5.4. Applying multiple performance constraints
• 10.6. Rejection
• 10.6.1. Adding reject option to a trained pipeline
• 10.6.2. Reject curve for distance-based rejection
• 10.6.3. Reject curve for rejection close to the decision boundary
• 10.6.4. Adding reject option to a specific operating point
• 10.7. Estimating ROC with variances

• 11.1. Feature extraction
• 11.1.1. Principal Component Analysis (PCA)
• 11.1.2. Linear Discriminant Analysis (LDA)
• 11.2. Feature selection
• 11.2.1. Individual feature selection (feature ranking)
• 11.2.2. Random feature selection
• 11.2.3. Forward search
• 11.2.4. Backward search
• 11.2.5. Floating search
• 11.2.6. Initialization of the selection searches

• 12.1. Detectors
• 12.1.1. Setting the detector operating point via thresholding
• 12.1.2. Detector for the all data
• 12.1.3. Setting the detector operating point via ROC analysis
• 12.1.4. Visualizing detector decisions on the image data
• 12.2. Discriminants
• 12.2.1. Nearest mean classifier
• 12.2.2. Linear classifier assuming normal densities
• 12.2.3. Quadratic classifier assuming normal densities
• 12.2.4. Gaussian mixture models
• 12.2.5. k-NN classifier
• 12.2.5.1. Prototype selection
• 12.2.5.2. Using large data sets
• 12.2.6. Parzen classifier
• 12.2.7. Neural network
• 12.2.8. Naive Bayes classifier
• 12.2.9. Decision tree classifier
• 12.2.9.1. Growing tree without pruning
• 12.2.9.2. Controlling the tree pruning
• 12.2.10. Random forest classifier
• 12.2.11. Support Vector Machine
• 12.2.11.1. Linear support vector machine
• 12.2.11.2. Polynomial support vector machine
• 12.2.11.3. Grid search for sigma and C parameters
• 12.2.11.4. Multi-class support vector machines (RBF)
• 12.2.11.5. Accessing support vectors
• 12.3. Classifier combining
• 12.3.1. Fixed combiners
• 12.4. Hierarchical classifiers

• 13.1. Introduction
• 13.1.1. Example of using cluster analysis to build a pixel classifier
• 13.2. Clustering with sdcluster command
• 13.2.1. Clustering all data irrespective of classes
• 13.2.2. How to cluster already clustered data set?
• 13.2.3. Changing default cluster names
• 13.2.4. Removing cluster labels
• 13.3. Clustering algorithms
• 13.3.1. k-means algorithm
• 13.3.1.1. Obtaining cluster labels with sdkmeans
• 13.3.1.2. Accessing prototypes derived by sdkmeans
• 13.3.2. k-centers algorithm
• 13.3.3. Gaussian mixture model

• 14.1. Introduction
• 14.1.1. Algorithm function
• 14.1.2. Algorithm training and execution
• 14.1.3. Converting algorithms into pipelines for out-of-Matlab execution

• 15.1. Introduction
• 15.1.1. Cross-validation by rotation
• 15.1.2. How are the errors computed
• 15.1.3. Setting random seed
• 15.2. Accessing algorithms trained in cross-validation
• 15.3. Accessing per-fold data sets
• 15.4. Cross-validation by randomization
• 15.5. Leave-one-out evaluation
• 15.5.1. Leave-one-out over property

• 16.1. Introduction
• 16.2. Execution of classifiers with command-line sdrun utility
• 16.2.1. Displaying pipeline information
• 16.2.2. Executing classifier on a data file
• 16.2.3. Executing classifier on samples provided in a string
• 16.3. Classifier execution in Microsoft Excel worksheets
• 16.3.1. Execution on data stored in a different workbook
• 16.4. Executing classifiers from LabView
• 16.5. Executing classifiers from Matlab/Matlab compiler
• 16.5.1. Loading a classifier pipeline
• 16.5.2. Executing a classifier on new data
• 16.5.3. Working with multiple pipelines
• 16.5.4. Removing pipelines from memory
• 16.6. Classifier embedding using C/C++ language API
• 16.6.1. Complete C application example
• 16.6.2. Using multiple pipelines
• 16.6.3. Handling decisions

• A.1. Pattern Recognition Books