Class 4: Content based recommendation engine using image data

What you need to do before the class

1. Please finish the Shirt Exercise in the previous class
2. Please use the Shirt Exercise pictures to outline a plan. Describe in detail how you want to convert the image into numerical representations. Please don't use deep learning for this task.

The level of details:

1. All steps
2. For each step, describe the packages and the functions in the packages you want to use.
3. Experiment processing the shirt images using your methodology.

Class content

1. Presentation: Present the Shirt Exercise solution. The one with manual feature abstraction. No image conversion to the numerical representation

2. Presentation: Present your methodology and experiment results of images to number conversion.

3. Exercise: use the shirts' numerical representation from the image to replace the manually created features in your code. See whether the result makes intuitive sense. Is it better? What if you combine with the two different kinds of features?

4. Exercise: Please go to MeetFresh Website to download the pictures of their products. Create manual features based on the pictures. Apply your content-based recommender code to this new dataset.

5. Ad hoc exercise: user profile generation through a questionnaire

   1. Use less than 10 questions to understand user's preferences
   2. Understand user's preference for each feature
   3. (stretch) Understand user's weighting of each feature
   4. (stretch) Able to get a non-binary number for user preference

   

---

Additional resources

OpenCV Website

OpenCV basics

pip install opencv