From Arno Schaepe’s experience

Turning a research image-analysis method into a core software product

Definiens had an image-analysis technology at prototype stage. Arno, who joined as a software developer in 1998, took it to the company's core product and was responsible for it for many years.

Industry
Health and life sciences
Organization
Definiens
Arno’s role
Software developer, later responsible for the core product
When
From 1998

The business need

Definiens, founded in 1994 by Professor Gerd Binnig, developed Cognition Network Technology for image analysis. The technology first became available as a product with the launch of eCognition in May 2000.

Definiens needed to take its image-analysis technology from an early prototype to its core product, and then keep that product stable as it scaled.

What Arno did

  1. Arno joined Definiens as a software developer in 1998.
  2. Arno took the company's image-analysis technology from an early prototype to its core product.
  3. In 2000 Arno co-authored, with Martin Baatz, a paper on multiresolution segmentation, an optimization approach for high-quality multi-scale image segmentation.
  4. Arno is a named co-inventor, with Gerd Binnig and Guenter Schmidt, on a Definiens patent for a cognition integrator and language (priority 2002), in which users define classes and process steps in a scripting language that link image pixels to objects.
  5. Arno then held responsibility for the product for many years, focusing on stability and scaling.

Technology and approach

  • Object-oriented image analysis that groups pixels into objects and sorts the objects into classes
  • Multiresolution segmentation
  • A scripting language for defining classes and process steps
  • Object-oriented image analysis for high-content screening of cells, described in a 2006 paper Arno co-authored on the Cellenger software

Outcome

  • Semantic Scholar lists about 2,750 citations for the 2000 multiresolution segmentation paper.

Decisions this experience speaks to

  • How do we take our research prototype to a working product in the market?
  • Should we use classical scientific computing and machine learning, or large language models, for our problem?

Related services and use cases

Sources

  1. Arno Schaepe’s profile on CTO Grand Rounds
  2. Cognition Network Technology - wikipedia.org (opens in a new tab)
  3. Multiresolution Segmentation (2000) - Semantic Scholar (opens in a new tab)
  4. US7873223B2 Cognition integrator and language - patents.google.com (opens in a new tab)
  5. Publications by Schäpe A - Europe PMC (opens in a new tab)