Research Group Prof. Dr. Kai Eckert

Global Coffee Atlas – AI-Based Recognition of Coffee Varieties to Support Smallholder Farming

Funded by the Konanz Foundation, this doctoral research project develops methods for the automated image analysis and classification of coffee plant varieties. The goal is to provide smallholder coffee farmers worldwide with an accessible tool for variety identification, serving as a basis for recommendations on adapting agricultural practices to climate change.

Background and Problem Statement

Climate change is increasingly affecting yields in coffee farming. While larger agricultural operations can respond to shifting conditions by adopting better-suited varieties, smallholder farmers frequently lack both the expertise and the financial resources to make such adjustments. A fundamental prerequisite for targeted adaptation strategies is the accurate identification of cultivated coffee varieties — a task that is morphologically challenging even for specialists, and prohibitively expensive at scale when relying on DNA sequencing.

Project Objectives and Funding

The doctoral project of Katharina Salewski, M.Sc., funded by the Konanz Foundation, aims to develop and scientifically validate AI-based methods for the automatic recognition and classification of coffee varieties. The work builds on a Master’s thesis in which an initial classification approach was developed from over 2,000 images taken at the Wilhelma Botanical Garden in Stuttgart, enabling the distinction of 20 Coffea arabica varieties.

Methodology

The project centres on the combination of AI-driven image analysis with a purpose-built knowledge graph that represents botanical characteristics in a structured form. This approach enables variety identification from photographs taken with standard smartphones. The method will be progressively extended to all approximately 100 coffee varieties held at the Wilhelma Botanical Garden and validated under field conditions, where plants may express differing phenotypes across environments. Validation against DNA sequencing results is planned.

Broader Context

The doctoral project is embedded within the larger Global Coffee Atlas initiative, which envisions a globally accessible, crowdsourced platform. Via a mobile application, farmers would be able to upload plant photographs and receive information relevant to their crops — including market prices, disease identification, and climate-adapted cultivation recommendations. The classification methods developed in this doctoral project form the technical foundation for that platform.

Team

The project is conducted as an interdisciplinary collaboration. Prof. Dr. Kai Eckert is the main supervisor for the doctoral dissertation and focusses on the AI and knowledge organization part.
Prof. Kirstin Kohler (TH Mannheim) provides further academic supervision, with expertise in user-centred design, computer science, and biology. Further contributors include Dr. Steffen Schwarz (coffee expert), and Dr. Björn Schäfer (Wilhelma Botanical Garden Stuttgart).