Computer Science Path

Bachelors’ Aera

It all started at the National and Kapodistrian University of Athens in 2020 when I began my Bachelor’s degree in Computer Science. From my very first lab in C, something sparked inside me. One of the assistants was using a unique editor to help me debug my code—VIM. There was something captivating about the way he worked, and I was instantly drawn to it.

As soon as I got home, I was on a mission. I dived deep into understanding VIM, and mastering its every command. It wasn’t just about the tool; it was about its sheer simplicity and elegance. VIM quickly became more than just an editor to me—it was a way of thinking, of coding with clarity and focus, free from the distractions of modern IDEs.

Five years later, after countless hours of writing C and C++ code, VIM is still my trusted companion. The bond I’ve formed with it is unbreakable. There’s no turning back, and honestly, I wouldn’t want to.

As I progressed, I discovered later a deep passion for Data Structures and Algorithms. By my second semester, this passion drove me to start coding fundamental Data Structures from scratch, even when it wasn’t required for the course. Curiosity led me to tackle the most challenging ones on my own, pushing myself to go beyond the standard curriculum. Early on, I began documenting my code with Git, and you can find some of my first implementations in the algorithms-data-structures repository.

In the years that followed, I devoted considerable effort to mastering C/C++, core software engineering principles (such as inheritance, and polymorphism), Linear Algebra, and Discrete Mathematics. The rigorous coding exercises at our university, primarily in C and C++, sharpened my understanding of the intricate mechanics behind core algorithms.

As I neared the end of my degree, it was about time to embark on my B.Sc. thesis. One day, after a long hours of lectures, I decided to attend a presentation on Robotics in our department. Little did I know, that decision would change everything. The speaker, Dr. Marios Xanthidis left a profound impact on me. He had completed his thesis at a different university, something quite uncommon in Greece. His journey and insights resonated deeply with me.

After his presentation, I reach out to him and share my growing interest in Robotics. That conversation opened a door I never expected. Xanthidis introduced me to his supervisor, Prof. Konstantinos J. Kyriakopoulos at the National and Technical University of Athens (NTUA). Following an interview, Prof. Kyriakopoulos placed his trust in me, allowing me to start my thesis at his lab, the Control Systems Lab. This opportunity marked the beginning of an exciting new chapter in my academic journey.

The journey was both challenging and immensely rewarding. It was during this time that I truly discovered my passion for Computer Vision and Deep Learning, and I began to understand how these fields could provide innovative solutions to complex robotics problems. For the first time, I saw my algorithms come into life, interacting with the 3D world through robotic platforms. Although I had no prior experience in these areas, I quickly immersed myself, learning new concepts in Robotics, Kinematics, ROS, Visual Servoing, and sensor technologies like the RGB-D Microsoft Kinect camera. I also delved deeply into Deep Learning techniques such as Object Detection, Segmentation, and Tracking.

Recognizing my rapid progress, particularly in Deep Learning, Prof. Kyriakopoulos entrusted me with the responsibility of peer-reviewing a paper for the 28th Mediterranean Conference on Control and Automation. It was an exciting challenge, especially given the mathematically intensive nature of the topic (Junction Detection for Drones via Clustering Techniques), but I committed myself fully, studying diligently to ensure a fair and impartial review. This experience was not only a testament to the growth I experienced during my thesis but also gave me a profound appreciation for the everyday life of researchers.

After completing my thesis, which you can find here, A Leader-Follower Mobile Robot Scheme using an RGB-D Camera and MobileNets, I finally developed a robotic system that utilizes an RGB-D camera and MobileNets for Object Detection, enabling a Leader-Follower scheme of mobile robots. The system was tested on Pioneer 3-AT and Summit-XL robots, demonstrating high accuracy and efficiency, with a processing speed of 19 Hz on a GeForce GTX1060 GPU.

Masters’ Aera

Upon receiving my Bachelor’s degree, I was eager to deepen my knowledge of Computer Vision and Robotics. In 2021, I began my M.Sc. in Computer Science at the Technical University of Munich (TUM).

The path was uncertain in a foreign country, but I embraced the unknown, believing that it offers great rewards when followed with instinct and courage.

During my studies, I focused on lectures that were highly theoretical, which helped me understand advanced methodologies. In my second semester, I wanted practical experience and joined the DevOps team at the Institute of Flight System Dynamics. There, I developed skills in Docker, Git, and Python while working on large projects and handling transitions between different repositories and technologies.

Later, I joined the Chair of Information-oriented Control at TUM and contributed to the SeaClear project, which became one of the most intriguing experiences of my academic journey. I implemented an MPC controller (designed by my supervisors) in C++ and Python to optimize a robot’s performance in clearing seabed trash.

In the same semester, I participated in the practical course “Cloud-Based Robotics in Deep Reinforcement Learning,” where my team performed exceptionally well. We extended our project into a research initiative and proposed a task-specific autoencoder (you can find our complete work at https://arxiv.org/abs/2309.11984). This project was a rollercoaster, as it was our first research endeavor. I vividly remember the final days before the deadline—we would trek through the snow to reach the lab, as it wasn’t directly accessible from the nearest train station. Despite the challenges, our motivation was incredibly high, and I’m deeply grateful to our supervisors, Mohammadhossein Malmir, and Josip Josifovski, for their unwavering support during this intense and unique period.

The same winter was particularly challenging, as I was also working as a student assistant while taking “Machine Learning,” one of the most mathematically intensive and highly regarded courses in my master’s program, taught by Prof. Stephan Günnemann. Many students chose to skip it due to its difficulty, but I couldn’t because of my passion for the subject. Despite the general advice from older students to avoid taking multiple courses simultaneously, I felt strongly about pursuing my interests. I took other courses, worked as a student assistant, and collaborated with Ludwig Gräf on our research paper. I’m not sharing this story to brag, but to inspire younger students. If you believe you can do something, you’re right, and if you think you can’t, you’re right too! Trust your intuition—it’s more powerful than you think.

In my fourth semester, I started working as a HiWi at the Chair of Robotics, AI and Real-time Systems as an ML and Robotics Engineer. I contributed to the Virtual Training Platform for Robot Learning project and had the opportunity to frequently operate real robots in the lab, including the LBR iiwa manipulator. This experience was highly fulfilling, as I spent a lot of time working with the LBR iiwa, a tall manipulator robot capable of lifting and handling objects. Despite the cold winter days in Munich, I found warmth in the rewarding work and the vibrant environment of the lab.

At the same time, I wanted to gain real-world experience in a company setting. I joined DeepScenario, where I learned about startup operations, focusing on rapid problem-solving and business acumen. One of the most interesting tasks I tackled was optimizing road network elevation profiles to align precisely with natural ground contours (ASAM OpenDRIVE), achieving an error tolerance of just 0.05 meters.

In October 2023, I started my M.Sc. thesis on Depth Prediction and SLAM under the supervision of S. Laina, S. Schaefer, and Prof. Dr. Stefan Leutenegger at the Smart Robotics Lab. I chose this thesis strategically, aware that it would push me in areas where I had limited experience, such as SLAM, optimization, and uncertainty estimation. The title of my thesis is “Scale-Aware Landmark-based SLAM in Canonical Space,” and we are currently working on publishing our findings.

Through these experiences, I discovered my true passion for Computer Vision, especially State Representation Learning. This field captivates me, as it involves developing robust latent embeddings or intermediate representations to solve the original task from a different perspective.

To any young student reading this: embrace experimentation and take risks. Don’t shy away from failure; each attempt is a step toward growth and discovery. You fail only when you decide to give up. There were moments when some of my supervisors doubted me (rightly so, as I wasn’t always experienced for the next challenge, though I was always willing to take the following step), but persistence helped me achieve my goals. Even if “failure” comes, you won’t be the same as when you started. In today’s world, where perfection is often expected, remember that novelty and creativity thrive on experimentation. Be willing to make numerous mistakes, learn from them, and take pride in the journey.