AI Tool 'Hespi' Revolutionizes Access to Herbarium Data, Accelerating Botanical Research

The Challenge of Herbarium Digitization

Herbaria worldwide house over 395 million plant and fungal specimens, forming an unparalleled record of Earth's biodiversity over time 1

1

. However, accessing this wealth of information has been a monumental challenge. Institutions have been striving to digitize their collections by photographing specimens and converting label information into searchable digital data. This process has been slow and labor-intensive, with many collections not expected to be fully digitized for decades.

Introducing Hespi: An AI-Driven Solution

Researchers have developed a new AI-driven tool called Hespi (herbarium specimen sheet pipeline) to address this challenge 1

1

. Hespi is open-source software that combines advanced computer vision techniques with AI tools such as object detection, image classification, and large language models.

How Hespi Works

The AI pipeline processes herbarium specimen images through several steps:

1. Text extraction using optical character recognition and handwritten text recognition
2. Error correction using OpenAI's GPT-4 Large Language Model
3. Identification and extraction of key information such as taxonomic names, collector details, location, and collection dates

Hespi can process a specimen sheet in seconds, converting the information into a digital format ready for research use 2

2

.

Accuracy and Potential Impact

The researchers tested Hespi on thousands of specimen images from various collections worldwide, achieving a high degree of accuracy. This tool has the potential to save significant time compared to manual data extraction methods. A graphical user interface is being developed to allow herbarium curators to manually check and correct results 1

1

.

Beyond Herbaria: Expanding AI Applications

The potential applications of Hespi extend beyond herbarium collections. The researchers are collaborating with Museums Victoria to adapt the tool for digitizing the museum's fossil graptolite collection, comprising about 12,500 specimens 2

2

.

Furthermore, a new project with the Australian Research Data Commons (ARDC) aims to make the software more flexible, allowing curators in various institutions to customize Hespi for different types of collections 1

1

.

Transforming Biodiversity Research

The development of AI tools like Hespi represents a significant leap forward in biodiversity research. By mobilizing large volumes of specimen-associated data, these systems enable new and innovative applications at an unprecedented scale. AI has already been used to automatically extract detailed leaf measurements and other traits from digitized specimens, unlocking centuries of historical collections for rapid research into plant evolution and ecology 2

2

.

As AI continues to reshape various aspects of scientific research, tools like Hespi have the potential to transform access to biodiversity data, accelerating progress in fields such as ecology, evolution, climate science, and conservation.