Machine Learning Project on Image Processing

The topic is ML for Image Classification, which is applying machine learning (Do NOT use
deep learning methods) to classifying images. The images can be in the following types:
• Natural Images
• Medical Images (e.g., PET, CT, MR)
• Pathological Images
• Remote sensing image
• biometric image (e.g. Face images, Fingerprint image)
• Any other possible image types

You need to use one enhanced version machine learning algorithm and an ensemble learning
algorithm to implement image classification.
1. The enhanced learning algorithm means applying the new developed method based on
the original algorithm. This is the midterm goal. The enhanced algorithm can be found in
the literature.
2. The ensemble learning means building an ensemble learning system based on the
enhanced learning algorithm to perform image classification. This is the final goal.
3. The original machine learning includes but not limited to the following algorithms:
• KNN
• Decision Tree
• Naive Bayesian classifier
• Rule-based classifier
• Support vector machine (SVM)
• Artificial neural network
• Logistic regression
• Any other possible classifiers
4. The whole workflow should contain feature extraction (SIFT, bag of word, intensity,
geometry and texture features), feature selection, model construction, model testing and validation
(cross validation).

It is highly suggested to use open source datasets in your Database, instead of fake data created
by your own group. Here are some websites with public datasets but not limited to:
a. https://www.kaggle.com/datasets
b. https://www.cancerimagingarchive.net/
c. https://archive.ics.uci.edu/ml/datasets.php