Latent Diffusion U-Net Representations Contain Positional Embeddings and Anomalies: Analyzing popular Stable Diffusion models, revealing three phenomena: (1) the presence of a learned positional embedding in intermediate representations, (2) high-similarity corner artifacts, and (3) anomalous high-norm artifacts.
SCAM - typographic attack dataset: Introducing SCAM, a typographic attack dataset for evaluating the robustness of (large) vision language models. Additionally, we evaluate popular models on the dataset, showing their susceptibility to typographic attacks.
SD Representation Similarity Explorer: An advanced interactive visualization tool for exploring representation similarities in text-to-image diffusion models.
sdhelper: A Python helper package for working with Stable Diffusion models, enabling easy extraction of U-Net and transformer representations.
Discovering Interpretable Directions in the Semantic Latent Space of Diffusion Models: Exploring and visualizing meaningful directions in the latent space of diffusion models.
H-Space Similarity Explorer: A simple interactive visualization tool for exploring representation similarities in text-to-image diffusion models.
Rover (ERC 2023 & 2024): Building a rover at the student space club BEARS for the simulated Mars yard at the annual European Rover Challenge. In 2024, we won the 7th place of 27 teams and the best Navigation (Droning) and best Science (Geological Exploration) awards!
TrainPlot: A Python library for real-time visualization of ML training metrics in Jupyter notebooks.
NiceHTML: A simple HTML alternative demoing an intuitive and streamlined alternative syntax for web development.
Blog: Offline RL with Diversified Q-Ensemble: An in-depth exploration of state-of-the-art approaches in offline reinforcement learning.
py2math: A small Python utility that automatically converts Python objects into LaTeX mathematical notation.
Blog: Safe Training in Reinforcement Learning: A comprehensive, interactive guide to curriculum induction in reinforcement learning.
Simple Toy Language: An educational implementation of a programming language parser and interpreter.
Background
Interested in deep learning, including diffusion models, explainability, and representation learning.
Miscellany
Personal interests include building and researching, with a particular focus on the field of deep learning.