Bootstrapping ML-driven CT Analysis

Adoption of ML-driven CT analysis in inline settings is primarily hindered by a lack of quality training data. We will make data acquisition and model training as fast, cheap, and easy as possible.

Stay Updated

Idea

Analysis of CT scans will have to be fully automated in smart factories. Complex analysis will involve machine learning algorithms that go beyond traditional computer vision. The adoption of machine learning is primarily hindered by a lack of quality data:

  • Manufacturers don't want to intentionally produce defect components for data collection
  • Scan acquisition in manual settings is time-intensive and expensive
  • Labeling is cumbersome and must be performed by domain experts

We help to bootstrap ML-driven automatic analysis by:

  1. Training a rough prototype solely based on defect-free scans or a CAD model
  2. Leveraging that prototype to assist in inline inspection, and make defect labeling much faster and easier
  3. Once enough labeled defects are available, we train a reliable model that can be deployed in fully automated settings

Team

Alexander Wiener

Alexander Wiener

  • BSc Information Systems at TU Munich with a focus on computer science
  • 1 year of (part-time) software engineering experience
  • Co-founder and former head of marketing of a large student club in Munich
Nicolas Stellwag

Nicolas Stellwag

  • MSc Informatics at TU Munich with a focus on machine learning in 3D settings
  • 4 years of (part-time) software engineering experience
  • Bachelor's thesis on ML for CT metal artifact reduction

Get in Touch

We're currently looking for partners to evaluate feasibility. If our project sounds interesting to you, we are always happy to talk! Feel free to send us an email or book a call.

Send Email