Automatic Plant Cover Estimation with Convolutional Neural Networks
Authors:
Matthias Körschens,
Paul Bodesheim,
Christine Römermann,
Solveig Franziska Bucher,
Mirco Migliavacca,
Josephine Ulrich,
Joachim Denzler
Abstract:
Monitoring the responses of plants to environmental changes is essential for plant biodiversity research. This, however, is currently still being done manually by botanists in the field. This work is very laborious, and the data obtained is, though following a standardized method to estimate plant coverage, usually subjective and has a coarse temporal resolution. To remedy these caveats, we invest…
▽ More
Monitoring the responses of plants to environmental changes is essential for plant biodiversity research. This, however, is currently still being done manually by botanists in the field. This work is very laborious, and the data obtained is, though following a standardized method to estimate plant coverage, usually subjective and has a coarse temporal resolution. To remedy these caveats, we investigate approaches using convolutional neural networks (CNNs) to automatically extract the relevant data from images, focusing on plant community composition and species coverages of 9 herbaceous plant species. To this end, we investigate several standard CNN architectures and different pretraining methods. We find that we outperform our previous approach at higher image resolutions using a custom CNN with a mean absolute error of 5.16%. In addition to these investigations, we also conduct an error analysis based on the temporal aspect of the plant cover images. This analysis gives insight into where problems for automatic approaches lie, like occlusion and likely misclassifications caused by temporal changes.
△ Less
Submitted 28 July, 2021; v1 submitted 21 June, 2021;
originally announced June 2021.
