Programming Assignment #3

Project Description

The goal of this project is to solve an indoor vs. outdoor image classification problem using machine learning. Starting code:

• classifiers.py Your classification code will go here. You should not change the method signatures for any of public methods.
• evaluate.py This is code for testing your classifiers. You probably won't need to make any changes to this code.

Training and testing Data:

• images.zip This should be unzipped in the same director as the Python files above.

You should be able to test the provided classifiers by running

$ python evaluate.py

from the terminal. Any new classes that you add to classifiers.py should be tested automatically.

Requirements

You will need to apply at least two different classification algorithms to this task, each of which must be tested with at least two different methods for generating features. This means you will need to create at least four new subclasses of the ImageClassifier base class.

One of the classification algorithms that you test must be the logistic regression algorithm that was implemented for HW6. The second can be anything you think will work well.

One (crude) algorithm for generating features has been provided. You will need to develop at least one additional approach that is more sophisticated.

You do not need to code your second classification algorithm from scratch. There are several well-developed Python machine learning packages. I suggest that you take advantage of one of those. Possibilities include:

• scikit-learn
• MDP
• mlpy

If you want to use something other than these three, please talk to me first.

At least one of your classifiers should have an accuracy significantly higher than 69.45% correct.

You must also submit a short (1-2 page) report describing your approach to the problem and your results. Your report will be graded on the basis of content as well as style. Your writing should be clear, concise, well-organized, and grammatically correct.