Category: Uni Projects

No comments

Reinforcement learning agent with visual inputs

  By , ,

As the module “individual profiling” in university, I created reinforcement learning agent working only on visual inputs, which could generally control anything on the computer. It was mainly built to play a certain video game, but can (in theory) generalize to do anything with visual input. It just needs an interface class to be written which converts the outputs to the desired thing to do.

For more information, visit the GitHub repo of the project (there is also in-depth documentation of the creation of the project).

Here are the two main examples used to show the best progress in two different games:

1. Driving nightmare (a game jam game created by a team of three people including me)

    This diagram shows the learning progress over 1600 iterations. The green line representing how long each run was (the higher the better) with the yellow line showing the average. The “loss” in blue being how far off the model thinks it is from the expected result.

    2. A simple Flappy bird like program built specifically for the AI. The flappy bird game can be advanced by code in specific steps, so it can wait for a slower working network without dropping any inputs.

    No comments

    (Math:) Projective space visualizer

      By , , ,

    In my university module “Higher Mathematics” we learned about projective space and the different representation with the hemisphere. (Or as we lovingly called it: “salad bowl”) This was part of the basics to understand Elliptic-curve cryptography.

    Since I had a hard time wrapping my head around the secondary representation of projective space, I decided to create a visualizer in the game engine I was familiar with at the time: Unity3d. I built a very crude module that can creates a 2D plane with a translucent mesh which can morph between the two representations. And I also added the intersecting lines together with the points to visualize where the points are at all times. This helped me understand the topic more deeply and didn’t take too much time to make.

    Since it wasn’t planned to be a finished project, it is not really polished, but you can still find the working unity project on the GitHub repository: https://github.com/MisterIXI/projective-space-visualization

    No comments

    Bachelor Thesis: Non Destructive Reverse Engineering of PCBs

      By , , ,

    Full title of the thesis: Non Destructive Reverse Engineering of Printed Circuit Boards using
    Micro Computed Tomography and Computer Vision

    This post only aims to illustrate the main contents of the bachelor thesis, for the full overview, the original paper is best read in it’s full form.

    Here is the abstract of the thesis:
    Reverse engineering (RE) of printed circuit boards (PCBs) is used for a variety of purposes, such as in computer forensics and quality assurance. Usually RE is very labor-intensive or destructive, since it pertains either manually measuring all visible contacts, including desoldering the components for the covered pads and mapping them out individually, or the process is done by milling away layer by layer to see inside the object and uncover the traces. This thesis aims to automate the process as much as possible while being non-destructive. To achieve this, micro computed tomography (µ-CT) will be used to scan the PCB while information will be extracted with the help of computer vision.

    The thesis researches the possibilities of using x-ray to reverse engineer PCBs. This makes it possible to understand PCBs without the need of damaging them using different methods.

    The program was not finished at the end of the thesis, since the reconstruction part was still missing, but the whole procedure was shown to work in theory. Here are a few pictures taken from the thesis to visualize the problems:

    Left to right: CT scan, pre-processed CT scan, edge detection visualized, original picture

    This is a comparison of the fix by tilting the PCBs when scanning in a certain way:

    This picture shows the edge detection up close and explains the coloured lines:

    The picture below shows the algorithm recognizing two traces on the PCB

    No comments

    Mixed reality online multiplayer board game simulator in Unity 3D

      By , , ,

    Together with a teammate, I created a multiplayer MR experience in which you can play Chess and Go. This was for the module “Windows App development” Used for this project was Unity3D, MRTK and PUN2. Check out the description and code on the GitHub repository.

    The project was made to work on an Augmented Reality device such as the Microsoft HoloLens and Virtual Reality Headsets such as the HP Reverb. The idea was that two players could connect with either platform and play with each other. Initially there were plans to integrate the table recognition of the MRTK for the AR devices, but that was scrapped due to time constraints.

    In the end, project had the following features:

    • peer to peer Online Multiplayer
    • AR and VR support
    • Builtin Chess and Go modes (no rules, just board and figures)
    • import of custom games with board texture, 3d models, snap positions, etc.
    • control with hand gestures for AR (Hololens)
    No comments

    Note analyser

      By , , , , ,

    As a university project for the module “Multimedia”, I was in a team to develop hardware which would recognize single notes or cords being played on an instrument (tuned to piano sounds).

    The project was run on a Raspberry Pi, and used the Fast Fourier Transformation to get the information needed. For a full writeup you best checkout the GitHub repository and the Printables entry.

    No comments

    Dodge-Box – a learning/university project

      By , , ,

    Dodge-box is a university project for the module “Software Engineering” in which we were supposed to do some sort of programm in a group with version control. We, as a group of three, decided to do my go-to learning project “Dodgesquare” in javafx.

    We didn’t have all too much time, so the project ended up pretty crude, but it works and we got a good grade.

    GitHub repo: https://github.com/MisterIXI/dodge-box