Image recognition as a tool for museum collections

In the project Automatic image recognition as a tool for museum collections, Naturalis and several other Dutch museums are investigating the use of artificial intelligence for object registration in museums. In this pilot machine learning is used to train several models on scans or photos of specimens, so that they learn how to identify objects.

Summary

In this project, we are developing AI models for different types of collection. We also investigate for which types of collections AI can and cannot be used, and what is needed to develop a reliable AI for this purpose. Museums, both in the Netherlands and abroad, that lack the expertise to identify species by themselves, might benefit, too. Machine learning may be able to help them identify the objects in their collection. The first models can be found here (page is in Dutch only, currently)

Subprojects

The focus is on five subcollections in several museums. We develop models for recognizing butterflies, bird eggs, cone snails, masks from different cultures and the eggs of rays and sharks.

Hidden moths: Make papillots visible in a streamlined way

In the papillot project, non prepared moths, which have been stored in pouches for decades, are made accessible to science by photographing, digitizing and storing them in a well-organized manner.