The goal was to create a classifier. We choose to make a painting classifier because there are a lot of extensions possible. We made a Neural Style Transfer (NST) as an extension. A NST makes it possible to transfer the style of a painting to another image. The application (only the classifier) is hosted on DigitalOcean.

I was responsible for the NST part and hosting the application.

The goal of this project was to use the learned techniques from the class. We decided to use fast.ai and OpenCV. First, we analyzed the problem to get a better understanding how we were going to implementing this AI. To collect data that we could use to train we played the game and saved a screenshot of the screen and the input of the controller every second. Since our data existed of an image and the controller input, we chose to use regression. The AI predicted the controller input when given an input frame.

I was responsible to collect and check the data to make sure there weren't too many anomalies.

Below you can see a clip of the final product.

The goal was to develop an app for Herentals that would make it more efficient to collect all the garbage bins. The team consists of students of different majors. Because of this we had a multidisciplinary team, some worked on the hosting, some on the application, others on the electronics. I worked on the visualizations. We had some sensors that measured how full the garbage bins were, how much they weigh and if there was a fire. These sensors sent this data to a SQL database. To get the data on the application we used a .NET backend to send it to the Angular frontend.

I was responsible for the data visualization of all the garbage bins. These visualizations are made with Qlik Sense.

You can see my part from 11:13 to 14:52 in the video bellow.

The goal of this project is to make it easier to organize staff parties. I made it possible to easily add activities to do during the day. It's also possible to invite guests by e-mail. The guests get an e-mail with a personal link with which they can register for the party. Here they can select which activity they prefer and in which parts they want to participate. I had multiple meetings with the client to make sure the functionalities and designs were implemented as expected.

I learned how important it is to communicate well within a team. It was basically a solo project for me since my teammates didn't do anything. They also didn't communicate to me if they were planning to help.

The goal of this AI is to answer questions. It learns these answers by looking through multiple documents and searching the answer in them. When it gets a question it which documents it can best check to get the answer. This is done with: tf-idf.

Trough this project I learned more about informaion retrieval and NLP in general.

The goal of this AI is to model all relationships inside a family to form a Bayesian Network. When you have this network and have the probabilities of the disease you can calculate the probability that someone inherits a disease from their parents.

I made this project because I'm interested in genetics and AI in the medical field.

The goal is to implement the Q-learning algorithm. The algorithm plays a lot of games against itself and updates every time the values of a Q-table. Because of this the AI gets better over time. After some time, it's impossible to win against the AI as a human because it got so good.

A little demonstration:

The goal of this AI is to solve a crossword using the AC-3 algorithm. This algorithm checks if every node in the network is arc consistent. It backtracks all the nodes to check that every node has the correct word in it, and they are all different.

The goal is to implement an AI that uses knowledge representation and reasoning. This means that the AI 'knows' the current state of the world and solves complex problems because of this information.

A little demonstration:

The goal is to implement the minimax algorithm. This algorithm makes it impossible to win as a human. It looks at every possible move it can make and calculates which one is the best.

A little demonstration:

The goal is that people can buy and sell stocks. They can see how much their stocks are worth now. It uses a SQLite database to save everything. To fetch the stock data, I used the IEX cloud API.

The goal is to better understand how filters work and implement them myself. This is done by editing the pixels themselves.

You can give the application a whole book to spellcheck. Below you can see how long it took to check all the words in the movie 'La La Land'. The hard part is when you check large texts because they can take a lot of time if the application isn't efficient. That's why I chose to use a hash table where every value is linked to a linked list.

I learned how important it is to write efficient programs. I also learned how linked lists work and how to implement them.