HW3 Naive Bayes
Due Oct 25, 11:59pm

In this homework you will train a Naive Bayes classifier on word and picture features.

Data

We will again use shopping images with associated text descriptions, this time for 4 bag related categories (hobo, shoulder, clutch, totes). Download the data here.
Your training data for this assignment will be for each bag category, those images numbered 500 and below. Your testing data for this assignment will be for each bag category, those images numbered above 500.

Part 0 - Background Reading

Read and understand Visual categorization with bags of keypoints by Csurka et al. We will be implementing something similar, but using both image and word features in our classifier. Also read descripion of Naive Bayes classifiers here and lecture notes for Oct 11 here.

Part 1 - Image Representation

We will model images as a histogram of visual word counts. Our visual word code book will be found by clustering shape based image features.

Part 2 - Text Representation

We will model text descriptions as a histogram of word counts.

Part 3 - Training Classifiers

In this part of the homework you will train your Naive Bayes classifier. This means calculating P(F_i|C_j) for each feature_i and each category_j. You can assume that P(C_j) is uniform over the 4 categories.

Part 4 - Image Classification and Confusion

Here we will classify test images using our trained classifier.

What to turn in

Hand in via email to cse595@gmail.com: