2022 was a great year for me. I like to do small summaries every now and then to make the progress more visible to myself and get motivated. Here’s the key takeaways of my 2022 as well as my plans for 2023.

A new job: Basemark

I joined Basemark in January. My title was “computer vision engineer” or something like that. My first task was to create benchmarks for ML-based computer vision algorithms on SoCs.

I have developed the following AR application, benchmarked it on several SoCs and GPU-accelerated it using TensorRT.

Markdown Monster icon

I also participated in developing BATS.

Creating a blog and a website

I actually created this website in late 2021. Most of the posts and the invisible background work happened during 2022. I had a dream of collecting and combining all the cool and visible results of my projects in a single place. I found out that creating a neat webpage is a bit of work but I learned the basics quite fast. In general getting projects ready and creating polished reports out of them is difficult. I’ve enjoyed the world of web development quite a lot and it’s mostly new to me as my tech stack is leaning more towards real-time compute, c++, desktop apps, ML and graphics.

I still think Jekyll is an awesome tool to manage content, publish easily and make beautiful webpages. It’s the perfect combination of techyness and ease of use compared to Squarespace or Wordpress for example. I still get to do commits, combine markdown and html and have the customizability of code.

Music

I did my first full length solo guitar concert in June 2022. I basically wanted to prove myself that I am capable of practising 45 minutes of music and performing it. It was possible but challenging.

I chose music by Leo Brouwer, Johann Sebastian Bach, Isaac Albeniz to name a few composers and even played one of my own original compositions. Needless to say that I had to practise a lot, take a few guitar lessons and do a few small warm-up concerts at the old people’s home.

I am also creating the soundtrack for the video game Twilight Ferry. It is a Kalevala-inspired visual novel that tells the story of the ferrymaiden in the land of the dead. More on that later.

Machine learning

I’ve done a lot of ML in 2022. Basically I’ve studied the whole “Nvidia” stack: CUDA, cuDNN, TensorRT, Nvidia hardware such as RTX GPUs, Xavier, Nano and Orin. It’s been fun.

I have completed about half of the course CS231N to deepen my knowledge of computer vision. I have read a hundred github repos on different models and cloned, installed and tested dozens. I have created an own classifier for images.

In the beginning of the year I had this question in my mind: “What is ML in the context of computer vision?”. I have spent this year trying to answer that question. I’m still not finished with it but I’ve made some progress.

Bicycling around the world

This year’s treat was a bicycle tour from Helsinki, Finland to Cluj-Napoca, Romania via Estonia, Latvia, Lithuania, Poland, Slovakia and Hungary. The tour gave me lots of insight and motivation.

Books

I love books. My favorite books of 2022:

  • Project Hail Mary by Andy Weir
  • The Count of Monte Cristo by Alexandre Dumas
  • Kon Tiki by Thor Heyerdahl (it’s the third or fourth time I read this)
  • Dark Matter by Blake Crouch

Coding projects

DooT2

We started researching our own doom bot. This is a long project that started maybe in 2020 or 2021. We are basically trying to create a learning agent that is capable of playing Doom2 maps by only screen pixel inputs.

See more here, here and here.

Obstacle Avoidance

I did my master’s thesis on obstacle avoidance for industrial robots. I wasn’t happy of my results back then and wanted to improve them so I created a new project to test obstacle avoidance and path planning algorithms.

Reinforcement Learning

I have been studying this quite a lot. While I have been studying materials such as Sutton and Barto’s legendary book or OpenAI’s Spinning Up -series I have been implementing the algorithms on Python. In general I think that you can’t learn by only reading. Playing with the ideas and extending them will help understanding the topic in-depth.

Web scraping

I’ve been studying what scapers are and how they can be used to get information from the interwebs.

OpenGL and computer graphics

This is a neverending story. Computer graphics and OpenGL has been hunting me for a long time. I know the basics and I’ve taken a few courses on the topic but I never built an intuition around the subjects. So I started studying graphics once more in pursuit of the joy of learning and understanding.

Software synthesizers

I’ve done two projects on this. Nothing cool to mention yet. It’s quite easy to do soft synths as there are much more resources now than there were 10 years ago.

Cleaning up my github repos

This is something we need to do every now and then.

2023

Wow, a whole year to do stuff. What should I do with it?

  • Play another guitar concert in June
  • Create ML models at work for autonomous vehicles
  • Continue creating the soundtrack for the video game Twilight Ferry
  • Create a bot that plays Doom2
  • Read a lot of books (mostly fiction)
  • Study interesting topics such as OpenGL, digital audio, 3D modeling and web developing
  • Take part in the open source community by opening PRs and testing code

That’s it. I better start coding.