In the summer of 2025, I experimented with blending competitions and courses into a single learning experience, aiming to discover whether playful problem-solving could be combined with structured teaching to create a more engaging way of learning. I had a lot of things to learn as well, and am writing this to share my experience.
The Competition
First, let’s look at the Decode the Drawings[1] competition. The challenge was to reconstruct drawings from short videos I recorded using a pen, with a small camera mounted on top (see Figure 1).
Solving this requires a mix of computer vision, geometry, and positioning techniques. The problem is complex, but it can be broken down into smaller parts, letting participants progress step by step toward a full solution. It resembles marker-based positioning used in AR[2], robotics[3], and motion capture[4], but with key differences: too few markers, size variations that matter, and critical cylinder crossings. These distinctions force participants to rely on first-principles thinking rather than prebuilt solutions (and LLMs currently cannot solve it adequately).
The competition ran for 2 months on my Discord server (that can be found here) in a public channel called #decode-the-drawings. All discussions and source codes are there, and my evaluation is also public. This setup let everyone track each other’s progress and see leaderboard shifts unfold in real-time. I also posted video updates[5] highlighting the key moments, announcing new prizes and the final results.
Community and Learning Dynamics
The competition was exciting on several levels. At first, it felt like a puzzle hunt: what are these hidden drawings, and how do we figure them out? Later, once people got working systems, it became more about optimization and fine-tuning, which made for some suspenseful leaderboard shifts (see Figure 2).
The winner was the participant whose reconstructions came closest to the real drawings. Runner-ups were decided through a badge system, where each badge represented a different technique (more techniques tried, more badges gained). In this way, exploration was rewarded alongside raw performance.
There were participants ranging from high-school students to IT professionals, and the winner turned out to be in the first group! Participants were from every continent (except Antarctica) so the competition helped promote Karelia to talented people all over the world. But the true highlight was the community that grew around it. The open discussion, shared code, and peer feedback created a learning ecosystem much richer than traditional lectures could offer. Participants could directly compare different problem-solving strategies, see each other’s thought processes, and learn from both clever insights and failed attempts. It felt very much like a work environment, so younger participants got a chance to experience this aspect as well. The community provided encouragement and accountability, both of which help participants persist through challenges they might otherwise abandon.
A total of 73 people participated, though most engaged only minimally: asking questions, proposing ideas, or sharing quick prototypes. Many of these small interactions were private messages I received via platforms I used to popularize the event: Discord, LinkedIn, Twitter, Facebook and even BuyMeACoffee. But most of the relevant discussions were on my Discord server (that can be found here) in the public #decode-the-drawings channel. The discussion produced 1,228 messages. Eleven participants showed sustained engagement (10+ active days). The peak activity occurred in May with 454 messages, coinciding with the initial problem exploration phase, followed by July with 427 messages during the optimization and fine-tuning phases. At its peak, daily activity reached 132 messages (evidence of intense collaborative problem-solving).
There was one noticeable drawback. For newcomers, seeing how far others had already progressed could be discouraging. This is probably why most competitions keep standings hidden until the deadline. But that format is designed with just scoring in mind, Decode the Drawings was always about learning first.
From Competition to Course
As the organizer, I got a huge benefit from this format. Seeing the problem through the participants’ eyes helped me to refine the related Computational Problem-Solving[6] course so it is at the right “entry level”. Many ideas from participants found their way into the course, shaping a narrative where solutions are built step by step, without assuming prior expertise. Teaching this way is surprisingly difficult once you’re deeply familiar with a topic; one tends to forget the learning process and treat concepts as self-evident. So, the competition was essential in shaping the course into its current form.
The course teaches how to break a problem into smaller parts (first-principles thinking), how to prioritize them, and how to isolate interconnected pieces so they can be worked on independently and checking whether these partial solutions are correct. We build a method from the ground up using simple reasoning and high school-level math. We first solve the problem by hand and learn how to automate it (using JavaScript). The algorithms are implemented in their simplest form and explained with plenty of visuals to walk through the logic. The target audience is beginners, especially high school students, who want practical ways to connect classroom math with real-world problem-solving.
Because it grew out of the competition, the course remains hands-on: learners code along and experiment with different approaches in their homework. The course does teach a complete solution to the problem, but the homework assignments, if implemented properly, will give a better result. So, in a sense, the competition is still open for those who want to improve their skills even further.
During the course, I also connect the concepts from this specific problem to tackling other real-world challenges. I did this in the hope that it affects student motivation when they realize that studying this one problem makes them competent in other fields as well (see Figure 3).
The course is currently ongoing and is offered to everyone interested in Open UAS[7]. So far, it has been well received judging by the view count and YouTube comments. I believe that because I discuss what happened during the competition in the video, participants have an interest to watch it as well, despite becoming really familiar with this problem over the past months.
In the end, organizing a competition-course combo like this brings significant value to everyone involved. It transforms learning into something playful and motivating, strengthens problem-solving skills, fosters a supportive community, and leaves a lasting impression on both learners and the organizer. It brings experts and novices together in a shared challenge, reframing education as both collaborative and experimental. By combining competition with structured teaching, it equips participants not only with technical know-how, but also with confidence and a resilient problem-solving mindset for the challenges ahead.
Some participant comments demonstrate the learning effectiveness of the experience:
- “I had never heard the term computer vision or any related term before this. Now I know a lot.” – exploreinfinity (19-year-old)
- “For me it was a joy to participate sort of like recreational and because i haven’t done nor learnt anything like this before it was educational.” – iyusuf.
- “I did learn a lot and had a lot of fun throughout the process.” – dannyfromdenmark
- “You literally made me learn Computer graphics in last for days … (but learnt lot of amazing things in last 4 days)” – tesseract4994
Writer:
Radu Mariescu-Istodor, lecturer, Karelia University of Applied Sciences
References:
[1] https://youtu.be/bZ8uSzZv0ew
[2] https://en.wikipedia.org/wiki/Augmented_reality
[3] https://en.wikipedia.org/wiki/Robotics
[4] https://en.wikipedia.org/wiki/Motion_capture
[5] https://www.youtube.com/watch?v=VuKX3JvlD8U&list=PLB0Tybl0UNfbCAHmvCpHZwL2sxo8sYrh-&index=4
[6] https://youtu.be/wH6gZB7m22k
[7] https://ella.eduplan.fi/karelia/computational-problem-solving/682722ead4168f741689894f