Tailoring convolutional neural networks for custom botanical data
Abstract Premise Automated disease, weed, and crop classification with computer vision will be invaluable in the future of agriculture. However, existing model architectures like ResNet, EfficientNet, and ConvNeXt often underperform on smaller, specialised datasets typical of such projects. Methods...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2025-01-01
|
| Series: | Applications in Plant Sciences |
| Subjects: | |
| Online Access: | https://doi.org/10.1002/aps3.11620 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832542875749974016 |
|---|---|
| author | Jamie R. Sykes Katherine J. Denby Daniel W. Franks |
| author_facet | Jamie R. Sykes Katherine J. Denby Daniel W. Franks |
| author_sort | Jamie R. Sykes |
| collection | DOAJ |
| description | Abstract Premise Automated disease, weed, and crop classification with computer vision will be invaluable in the future of agriculture. However, existing model architectures like ResNet, EfficientNet, and ConvNeXt often underperform on smaller, specialised datasets typical of such projects. Methods We address this gap with informed data collection and the development of a new convolutional neural network architecture, PhytNet. Utilising a novel dataset of infrared cocoa tree images, we demonstrate PhytNet's development and compare its performance with existing architectures. Data collection was informed by spectroscopy data, which provided useful insights into the spectral characteristics of cocoa trees. Cocoa was chosen as a focal species due to the diverse pathology of its diseases, which pose significant challenges for detection. Results ResNet18 showed some signs of overfitting, while EfficientNet variants showed distinct signs of overfitting. By contrast, PhytNet displayed excellent attention to relevant features, almost no overfitting, and an exceptionally low computation cost of 1.19 GFLOPS. Conclusions We show that PhytNet is a promising candidate for rapid disease or plant classification and for precise localisation of disease symptoms for autonomous systems. We also show that the most informative light spectra for detecting cocoa disease are outside the visible spectrum and that efforts to detect disease in cocoa should be focused on local symptoms, rather than the systemic effects of disease. |
| format | Article |
| id | doaj-art-1bdbd468e07b4c65848bdd0a203a785a |
| institution | Kabale University |
| issn | 2168-0450 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | Wiley |
| record_format | Article |
| series | Applications in Plant Sciences |
| spelling | doaj-art-1bdbd468e07b4c65848bdd0a203a785a2025-02-03T12:21:34ZengWileyApplications in Plant Sciences2168-04502025-01-01131n/an/a10.1002/aps3.11620Tailoring convolutional neural networks for custom botanical dataJamie R. Sykes0Katherine J. Denby1Daniel W. Franks2Department of Computer Science University of York Deramore Lane York YO10 5GH Yorkshire United KingdomCentre for Novel Agricultural Products, Department of Biology University of York Wentworth Way York YO10 5DD Yorkshire United KingdomDepartment of Biology University of York Wentworth Way York YO10 5DD Yorkshire United KingdomAbstract Premise Automated disease, weed, and crop classification with computer vision will be invaluable in the future of agriculture. However, existing model architectures like ResNet, EfficientNet, and ConvNeXt often underperform on smaller, specialised datasets typical of such projects. Methods We address this gap with informed data collection and the development of a new convolutional neural network architecture, PhytNet. Utilising a novel dataset of infrared cocoa tree images, we demonstrate PhytNet's development and compare its performance with existing architectures. Data collection was informed by spectroscopy data, which provided useful insights into the spectral characteristics of cocoa trees. Cocoa was chosen as a focal species due to the diverse pathology of its diseases, which pose significant challenges for detection. Results ResNet18 showed some signs of overfitting, while EfficientNet variants showed distinct signs of overfitting. By contrast, PhytNet displayed excellent attention to relevant features, almost no overfitting, and an exceptionally low computation cost of 1.19 GFLOPS. Conclusions We show that PhytNet is a promising candidate for rapid disease or plant classification and for precise localisation of disease symptoms for autonomous systems. We also show that the most informative light spectra for detecting cocoa disease are outside the visible spectrum and that efforts to detect disease in cocoa should be focused on local symptoms, rather than the systemic effects of disease.https://doi.org/10.1002/aps3.11620disease detectionmachine learningplant pathologyspectroscopy |
| spellingShingle | Jamie R. Sykes Katherine J. Denby Daniel W. Franks Tailoring convolutional neural networks for custom botanical data Applications in Plant Sciences disease detection machine learning plant pathology spectroscopy |
| title | Tailoring convolutional neural networks for custom botanical data |
| title_full | Tailoring convolutional neural networks for custom botanical data |
| title_fullStr | Tailoring convolutional neural networks for custom botanical data |
| title_full_unstemmed | Tailoring convolutional neural networks for custom botanical data |
| title_short | Tailoring convolutional neural networks for custom botanical data |
| title_sort | tailoring convolutional neural networks for custom botanical data |
| topic | disease detection machine learning plant pathology spectroscopy |
| url | https://doi.org/10.1002/aps3.11620 |
| work_keys_str_mv | AT jamiersykes tailoringconvolutionalneuralnetworksforcustombotanicaldata AT katherinejdenby tailoringconvolutionalneuralnetworksforcustombotanicaldata AT danielwfranks tailoringconvolutionalneuralnetworksforcustombotanicaldata |