Abschlussarbeit: Visual Pattern Recognition – Identifikation von Design-Clustern in der Zeitungsproduktion mittels spezialisierter Vision-Modelle
Core
Develop a specialized Vision AI system to automatically analyze and classify high-resolution print designs of local newspapers using Vision Transformers or LayoutLM to identify design clusters and support fully automated typesetting systems.
Role type
Research-focused Computer Vision Engineer (Thesis)
Builds
A prototype system that extracts design parameters (column logic, whitespace usage, font weighting) to identify newspaper design lineages.
Domain
Digital Journalism / Computer Vision / AI Research
Deliverable
production ML models
Required skills
Computer Vision, Vision Transformers, LayoutLM, Unsupervised Learning, Semi-supervised Learning, Python, Data Processing
Preferred skills
Experience with scientific research, benchmarking model architectures, comparing against LLMs
Technologies
Vision Transformer, LayoutLM, CNN, GPT-4o, Gemini
Responsibilities
Conduct benchmark of different model architectures, prototype an identifier for design patterns, evaluate model accuracy against leading Vision LLMs, validate system with unknown newspaper pages.
Seniority
Junior, Research Assistant
