Go Deeper Into Your Collection
Natural history collections hold an extraordinary amount of scientific knowledge, but much of it is still locked behind inaccessible handwritten and printed specimen labels. Transcribing the text is an important first step. But the resulting block of text still needs to be interpreted, structured, checked, and connected via knowledge sources before it can become truly useful. That is the challenge that DEEP by Picturae masters.
Research-Ready Data
DEEP by Picturae identifies the different elements on a specimen sheet, extracts the relevant information, and turns it into structured data. Scientific names can be checked against trusted knowledge sources such as Catalogue of Life and localities can be georeferenced. Collector names and numbers can be matched against an institution’s own registers or IPNI author names. The result is not simply a digital version of the label. It’s data that can be imported, searched, and used for research.
In the real herbarium sheet example shown, the locality was flagged for curator review, while the remaining fields passed the confidence threshold.
We Use Specialist AI Agents
Labels vary in language, handwriting, layout and age. That’s why DEEP by Picturae does not rely on one generic AI model to understand everything at once. Instead, specialist AI agents perform specific tasks. One detects the labels, barcodes, stamps, and annotations. Another reads printed and handwritten text. Others identify scientific names, collector information, and localities, checking the results against relevant knowledge sources.
A Plausible Answer is Not Necessarilya Correct Answer
No matter the type of natural history collection, you set the confidence threshold for your data extraction results. Anything that moves above the established threshold passes into the finished phase automatically, while those that do not will be sent to you directly for review. Your experts will approve or correct the data while checking the records against the original image. Both the original reading and the corrected value are kept, with a complete audit trail.
Accurate Interpretation Requires Context
A specimen is part of a larger, structured collection. It may sit inside a folder, which sits inside a box, or a drawer. Information known at box or folder level can be inherited by the specimens within it. When information conflicts, the record can be flagged instead of allowing the inconsistency to pass unnoticed. Institutions can also connect their own knowledge sources, including collector registers, georeferencing, and project-specific reference lists. This allows the platform to work with the knowledge that already exists around a collection, rather than asking AI to interpret every image in isolation.
Available Today as a Service
No matter who has digitized your specimens, they can be processed through DEEP by Picturae. Together, we configure the workflow around your collection, the fields you need, and the knowledge sources you want to use. We then process the material and coordinate the review of flagged records with your curators and experts.
Contact us to discuss how DEEP by Picturae can support your digitization project.
Data that Fits your Collection System
Extracting data is only useful if the result can actually move into the systems and workflows an institution uses. Every collection has its own requirements. The workflows of DEEP by Picturae are therefore configured per project. Institutions define the information they need. Relevant specialists and knowledge sources are selected. The resulting data can be aligned with standards such as Darwin Core and Dublin Core – or mapped to an institution‘s own schema. The aim is not to produce another isolated dataset that creates more work. It is to deliver structured records in a form that can become part of the collection’s existing data ecosystem.
Institutions Retain Ownership
AI processing raises important questions about data governance, data residency, and institutional control. For European institutions in particular, requirements around privacy, security, and the processing of data are important considerations when adopting AI technologies. DEEP by Picturae was developed with these requirements in mind, with an architecture designed to give institutions greater visibility and control over how their collection of data is processed and ownership of their images and the data produced from them.
Become a partner
If your institution has digitized herbarium specimens and would like to explore what our platform could unlock from them, now is a good time to start the conversation. Tell us about your institution and your digitized specimens, and we will get back to you promptly.
