RT Journal Article
SR Electronic
T1 Automation of tree-ring detection and measurements using deep learning
JF bioRxiv
FD Cold Spring Harbor Laboratory
SP 2022.01.10.475709
DO 10.1101/2022.01.10.475709
A1 Miroslav Poláček
A1 Alexis Arizpe
A1 Patrick Hüther
A1 Lisa Weidlich
A1 Sonja Steindl
A1 Kelly Swarts
YR 2022
UL http://biorxiv.org/content/early/2022/01/12/2022.01.10.475709.abstract
AB We present an implementable neural network-based automated detection and measurement of tree-ring boundaries from coniferous species.We trained our Mask R-CNN extensively on over 8,000 manually annotated rings. We assessed the performance of the trained model from our core processing pipeline on real world data.The CNN performed well, recognizing over 99% of ring boundaries (precision) and a recall value of 95% when tested on real world data. Additionally, we have implemented automatic measurements based on minimum distance between rings. With minimal editing for missed ring detections, these measurements were a 99% match with human measurements of the same samples.Our CNN is readily deployable through a Docker container and requires only basic command line skills. Application outputs include editable annotations which facilitate the efficient generation of ring-width measurements from tree-ring samples, an important source of environmental data.Competing Interest StatementThe authors have declared no competing interest.