Image Classification with SVM

In this project we're comparing the image classification performance of SIFT (Scale-Invariant Feature Transform), SURF (Speeded-Up Robust Features) and HOG (Histogram of Gradients). For training and testing we're using images from the Caltech-256 image dataset.

Contents

• img: Contains the training data
• matrices: Contains the training and test matrices with the images, and the features they contain
• src: Contains the source code

Prerequisites:

You'll need OpenCV and OpenCV Contrib 3.4.2.6 for Python:

pip install opencv-python==3.4.2.16 opencv-contrib-python==3.4.2.16

It's essential to use this version, since newer versions do not contain SIFT and SURF as they're patented.

Usage:

Parameters:

• Method: Choose feature extraction method: SIFT SURF or HOG
• Classes: Choose the names of the folder containing the image classes
• K: Choose the number of bins used for clustering (optional. Default: 100)
• Splits: Choose the number of splits (optional. Default: 3)
• Cross Validation: Set to True to use Cross Validation or to false for the opposite (optional. Default: True)

python main.py <sift/surf/hog> -c <image classes> -k <number of clusters> -s <number of splits> -cval <True for using CrossValidation>