Projects
Visual Transformer
A tiny experiment where I implemented a Vision Transformer (ViT) from scratch and compared it with a classic Convolutional Neural Network (CNN).
- 🔹 Explore how attention works under the hood
- ⚡ Compare speed and accuracy with CNNs
Probably, attention is not all you need 😅
Object Tracking
A demo project showing how to track objects in video using DINOv3 features. 🎥
- 🔖 Image understanding without labeled data
- 🔥 Robust feature embeddings
- 🎨 Smooth tracking heatmaps
Perfect for exploring self-supervised vision transformers and be impressed by modern AI 😎
TalkFast
A Telegram bot for learning German & English vocabulary with friends or family. 🎉
- 🎯 Practice new words in a fun way
- 🧠 Train your brain with timed challenges
- 👫 Play in teams with scoring & victory rounds
Whether you're a language learner or just want a fun party game, TalkFast is here to help 😁
Local Voice LLM Assistant
An experiment in running compact but capable language models fully offline on your laptop. A private AI assistant keeps all your data local — with no cloud and hidden limitations.
- 🔒 Full privacy and control over the process
- 🎤 Voice input for hands-free interaction
SwapFace
An app to swap two faces in a single photo using Laplacian Pyramid blending. Flexible CMake configuration allows you to enable:
- 🖼️ Work with single images
- 🎥 Work with videos
Stereo Breath Detection
Detect chest movements (breathing) in real-time using two cameras + OpenCL acceleration.
- 🎥 Stereo vision
- 📏 Disparity map → depth perception
- ⚡ GPU-accelerated via OpenCL