No Detail Found!

Integrating transfer learning in convolutional neural networks for bryophyte classification

Author(s)
Jessica Velasco Department of Electronics Engineering, Technological University of the Philippines,Manila,Philippines
Joshua Dela Cruz Department of Electronics Engineering, Technological University of the Philippines,Manila,Philippines
Karl Arthur Von Gomez Department of Electronics Engineering, Technological University of the Philippines,Manila,Philippines
Renz Jirecho Barquilla Department of Electronics Engineering, Technological University of the Philippines,Manila,Philippines
Dave Carlos Calixto University Extension Services, Technological University of the Philippines,Manila,Philippines
Benedicto Fortaleza University Extension Services, Technological University of the Philippines,Manila,Philippines
Abstract
This study delves into the rich and varied realm of bryophytes, which are non-vascular plants that are essential to terrestrial ecosystems. Over millions of years, bryophytes-which include mosses, liverworts, and hornworts-have, changed their surroundings by absorbing nutrients and water, releasing it gradually, and making room for new plant growth. The study explores the taxonomy of bryophytes, emphasizing the difficulties associated with manual classification because of the large number of species. The paper suggests using Convolutional Neural Networks (CNNs) for automated plant species classification to overcome these issues. Effective CNNs for bryophyte classification are DenseNet121, MobileNetV2, Xception, ResNet50, InceptionV3, and ResNet-152. The use of open-source datasets and the project's implementation utilizing Google Colab, the Keras platform, and a Tensor Flow backend are described in the methods section. Based on confusion matrices, loading times, and accuracy, the study evaluates the effectiveness of many CNN algorithms. Based on the results, MobileNetV2 is the best-performing model, with an average loading time of 31.30 seconds and 90.90% accuracy while ResNet50 has the lowest accuracy of 36.03% and the slowest model for the average loading time is Xception with 73.30 seconds. The accuracy and loading times of other models, such as ResNet152, and InceptionV3, differ. By automating the classification of bryophyte species, CNNs can be a useful tool for scientists and researchers, as this study concludes. The work advances knowledge about bryophytes and offers information on how machine learning might be used to classify them.
Publication Details
Page(s) 13-21
Published Journal: ARPN Journal of Engineering and Applied Sciences, Volume: 20, Issue: 1, Year: 2025

Request the full text from PASTIC

Keywords
Convolutional Neural Network, Python, Deep learning, bryophytes classification
References
Wäldchen J. and Mäder P. .2017 .Plant Species Identification Using Computer Vision Techniques: Systematic Literature review. Archives of Computational Methods in Engineering, 25(2) : 507-543.
Pawar M.,Shaha M. .2018 .Transfer learning for image classification. , : .
Segal T. .2022 .What is big data? Definition, how it works and uses. asp., : .
.2020 .. [5] Scientific research and big data, : .
Lübbecke M. .2015 .-data-saves-lives/index. CNN, : .
Gillis A. (n.d.). . .fbclid=IwAR1R7kRUswM3U4C0WqulUY1 F5VJv-1wQrX70KBlbpvghSxhEqRmi-tyQIhc.. TechTarget, : .
Yegulalp S. .2024 .-istensorflow-the-machine-learning-libraryexplained.html? fbclid=IwAR3gIRFO2OEf5R7g1XuPlBCwh1GSWM TXwaY4a_B3smzudU76MPrN2aWNcQQ.. InfoWorld, : .
Crooks V. .2021 .Bryophytes are tiny plants in a big changing world. , : .
Hornwort .1998 .| aquatic, submerged, liverwort. , : .
Lepp H. .2008 .What is a moss? Australian National Herbarium. , : .
Hepaticopsida .2022 .. , : .
Tsang S. H. .2018 .DenseNetdense convolutional network (image Classification). Medium., : .
Chen L. C.,Howard A.,Sandler M.,Zhmoginov A.,Zhu M. .1801 .Inverted residuals and linear bottlenecks. , : .
Rodriguez J. .2021 .-ofgoogles-mobilenet-architectures-to-improvecomputer-vision-models-ffb483ffcc0a.. , : .
. .Xception architectural design. (n.d.), : .
Akhtar Z. .2021 .Deep learning with depth-wise separable convolutions. OpenGenus IQ: Computing Expertise & Legacy, : .
. .The basics and a quick tutorial. (n.d.), : ResNet.
Boesch G. .2024 .- 2024 guide. It%20is%20an%20innovative %20neural,Residual%20Learning%20for%20Image% 20Recognition'., : .
T. A. N. ,Inception V3 Model Architecture. OpenGenus IQ: Computing Expertise .2021 .October 8. , : .
Shanmugamani R. . .~:text=Inception%2DV3%20w as%20proposed%20by,the%20ImageNet%20competit ion. , : 202014.
0

Citations

0

Downloads

27

Views