ME5405_Project

Project Execution

Before execute the project, please add all folders and subfolders to the MATLAB path, and then execute the corresponding files.

• For Chromosome Image1 (chromo.txt), the main function file is chromo.m
• For Characters Image2 (charact1.txt), the main function file is charact1.m

Classification Task

All 3 methods K-NN, SOM, and SVM are implemented, and the source code are placed in the folder Classification.

SVM

The dataset was transfered from .mat to .png with binarization operation.

run the file Classification/SVM/SVM_Char_recog.m, the corresponding result can be illustrated as:

K-NN

The main code of KNN-Classifier for execution is :

KNN_Run_this.m

• To run the KNN-Classifier, the paths of the training data are needed to be changed to your local path. The 'Sample' files are the set used to train while the 'TestingSet; is used to test. The output will be the accuracy.
• The second part applies trained KNN classifier to determine the characters we obtain in the previous task as rearranged.mat. The output will be the corrsponding label of each input.

  Noted: The label A B C are labeled as 4 5 6 correspondingly.

SOM

API:

• Sample1 ~SampleC : dataset (.mat) of each character(1, 2, 3, A, B, C).
• Testingset: validation set.
• mat_file_load.m: Matlab script to read and load the training set and validation set.
• SOM_training.m: Matlab script to train SOM classifier.
• SOM_validation.m: Matlab script to do the validation.
• SOM_confusion_matrix.m: Matlab script to visualize the confusion matrix.
• SOM_testing.m: Matlab script to test the recognization for characters in 'charact1.txt'.

To run the following files in sequence to realize the classification:

mat_file_load.m
SOM_training.m
SOM_validation.m
SOM_testing.m