Implement basic classifiers (Bayes', K-Nearest Neighbors, PCA, LDA) in MATLAB to achieve facial recognition.
For a formal definition of the assignment, please see the project description. For a summary of the results, please see my final report.
In order to keep the code readable and modular, each of the classifier and dimension reduction techniques have been implemented in separate MATLAB functions located in the classifiers directory. Helper functions are conveniently located in the utils directory. The main script is located in the top level of the code directory.
I have divided the script into the following sections:
- Preprocess Data
- Load Variables
- Divide Data (Training and Testing)
- Bayesian Classification
- K-Nearest Neighbors Classification
- Principal Component Analysis (PCA)
- Bayesian Classification after PCA
- K-Nearest Neighbors Classification after PCA
- Fisher's Linear Discriminant Analysis (LDA)
- Bayesian Classification after LDA
- K-Nearest Neighbors Classification after LDA
The main file contains initial conditions and parameters for different circumstances. To test different functionality, simply modify those state conditions variables.
The script can be run its entirety, or one section at a time to observe and analyze the results for the given section. At the end of the script, the results of each classification technique are displayed in a table. Below is an example table when running the whole script on the Face data set.
All input data can be found in the data directory in the form of .mat files. For any help or clarification, please review the help text to understand how each of the functions are meant to be used.
The extra credit script is written in the same format as the main script. In this script, I split the face data set 3 different classes corresponding to the states of "Happy", "Neutral" and "Illuminated". I used this data to train the classifiers to detect the state of a given subject, as opposed to recognizing the subject themselves.