Abstract
PremiseThe generation of morphological data in evolutionary, taxonomic, and ecological studies of plants using herbarium material has traditionally been a labor‐intensive task. Recent progress in machine learning using deep artificial neural networks (deep learning) for image classification and object detection has facilitated the establishment of a pipeline for the automatic recognition and extraction of relevant structures in images of herbarium specimens.Methods and ResultsWe implemented an extendable pipeline based on state‐of‐the‐art deep‐learning object‐detection methods to collect leaf images from herbarium specimens of two species of the genus Leucanthemum. Using 183 specimens as the training data set, our pipeline extracted one or more intact leaves in 95% of the 61 test images.ConclusionsWe establish GinJinn as a deep‐learning object‐detection tool for the automatic recognition and extraction of individual leaves or other structures from herbarium specimens. Our pipeline offers greater flexibility and a lower entrance barrier than previous image‐processing approaches based on hand‐crafted features.
Highlights
PREMISE: The generation of morphological data in evolutionary, taxonomic, and ecological studies of plants using herbarium material has traditionally been a labor-intensive task
Our pipeline offers greater flexibility and a lower entrance barrier than previous image-processing approaches based on hand-crafted features
With GinJinn, we provide plant scientists a tool for applying modern machine learning–based visual recognition to their own data sets without requiring a thorough theoretical background in machine learning and proficiency in programming, which is generally necessary to apply and deploy deep-learning object detection
Summary
GinJinn was originally developed as an internal tool for rapid iteration through deep-learning model architectures to find adequate neural network models for the detection and extraction of intact leaves in digital images of herbarium specimens for subsequent morphometric analyses It has since evolved into a general object-detection pipeline for the setup, training, evaluation, and deployment of bounding-box-based object-detection models with a focus on providing easy access to a high number of different model architectures with little manual work for the user, including the automated download of pretrained models if available. Our results indicate the applicability of training a deep-learning model for the detection of objects in preserved plant specimens that can potentially assist or even automatize the extraction of leaves from herbarium images or assist further annotations, even with a relatively small data set of only 243 images. This section should be considered a tutorial for new users
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
