By using the kNN approach and three similarity measures, build image classifiers

INSTRUCTIONS TO CANDIDATES

ANSWER ALL QUESTIONS

Assignment

In this assignment, we are going to use excerpts from the following datasets:

• AI Generated Faces from Generated.Photos https://generated.photos

• Turath-150K Image Database of Arab Heritage https://danikiyasseh.github.io/Turath/

• ANIMAL-10N Dataset https://dm.kaist.ac.kr/datasets/animal-10n/

Song, H., Kim, M., and Lee, J., "SELFIE: Refurbishing Unclean Samples for Robust Deep Learning" In Proc. 36th Int'l Conf. on Machine Learning (ICML), Long Beach, California, June 2019

 

The assignment is worth 30 points in total and is compromised of the following tasks. The classification scheme as well as the data can be found in the .zip file accompanying this portfolio. Please also use the attached template; if the template uses a different programming language than you want to use, please contact the module leader.

 

You can only use standard libraries that come with the language you have picked unless stated otherwise. For example, calling a kNN classifier from a scikit-learn package instead of implementing your own from scratch will yield 0 points.

 

Additional clarifications concerning this portfolio may be posted on the discussion board on Learning Central, so please remember to check it.

 

Task 1 [10] My first not-so-pretty image classifier

 

By using the kNN approach and three similarity measures, build image classifiers. You need to implement the kNN approach yourself, however, you can use libraries for any similarity measures (remember that some measures can make assumptions on the sizes of images etc.) You can assume that k=100 (if the code takes too long to run, feel free to decrease it to as low as k=10). You are allowed to use libraries to read and write to files, and to perform image transformations if necessary.

 

Task 2 [2] Basic evaluation

 

Evaluate your classifiers. On your own, implement methods that will output precision, recall, F-measure, and accuracy of your classifiers. Task 3 [6] Cross validation

 

Evaluate your classifiers using the k-fold cross-validation technique covered in the lectures (use the training data only). Assume the number of folds is 100 (if the code takes too long to run, feel free to decrease it to as low as 10 folds). Output their average precisions, recalls, F-measures and accuracies. You need to implement the validation yourself.

 

Task 4 [3] The curse of k

 

Independent inquiry time! Picking the right number of neighbours k in the kNN approach is tricky. Find a way you could approach this more rigorously. In comments, state the approach you could use, and provide a reference to it. The reference needs to be to a handbook or peer-reviewed publication; a link to an online tutorial will not be accepted.

 

Task 5 [6] Similarities

 

Independent inquiry time! In Task 1, you were allowed to use libraries for image similarity measures. Pick two of the three measures you have used and implement them yourself!

 

Task 6 [3] I can do better!

 

Independent inquiry time! There are much better approaches out there for image classification. Your task is to find one, and using the comment section of your project, do the following:

• State the name of the approach, and a link to a resource in the Cardiff University library that describes it

• Briefly explain how the approach you found is better than kNN in image classification (2-3 sentences is enough).

 

 

Attachments:

Related Questions

. Introgramming & Unix Fall 2018, CRN 44882, Oakland University Homework Assignment 6 - Using Arrays and Functions in C

DescriptionIn this final assignment, the students will demonstrate their ability to apply two ma

. The standard path finding involves finding the (shortest) path from an origin to a destination, typically on a map. This is an

Path finding involves finding a path from A to B. Typically we want the path to have certain properties,such as being the shortest or to avoid going t

. Develop a program to emulate a purchase transaction at a retail store. This program will have two classes, a LineItem class and a Transaction class. The LineItem class will represent an individual

Develop a program to emulate a purchase transaction at a retail store. Thisprogram will have two classes, a LineItem class and a Transaction class. Th

. SeaPort Project series For this set of projects for the course, we wish to simulate some of the aspects of a number of Sea Ports. Here are the classes and their instance variables we wish to define:

1 Project 1 Introduction - the SeaPort Project series For this set of projects for the course, we wish to simulate some of the aspects of a number of

. Project 2 Introduction - the SeaPort Project series For this set of projects for the course, we wish to simulate some of the aspects of a number of Sea Ports. Here are the classes and their instance variables we wish to define:

1 Project 2 Introduction - the SeaPort Project series For this set of projects for the course, we wish to simulate some of the aspects of a number of