Automated Machine Learning for High-Throughput Image-Based Plant Phenotyping
Automated machine learning (AutoML) has been heralded as the next wave in artificial intelligence with its promise to deliver high-performance end-to-end machine learning pipelines with minimal effort from the user. However, despite AutoML showing great promise for computer vision tasks, to the best...
Main Authors: | , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2021-02-01
|
Series: | Remote Sensing |
Subjects: | |
Online Access: | https://www.mdpi.com/2072-4292/13/5/858 |
_version_ | 1797395259957706752 |
---|---|
author | Joshua C.O. Koh German Spangenberg Surya Kant |
author_facet | Joshua C.O. Koh German Spangenberg Surya Kant |
author_sort | Joshua C.O. Koh |
collection | DOAJ |
description | Automated machine learning (AutoML) has been heralded as the next wave in artificial intelligence with its promise to deliver high-performance end-to-end machine learning pipelines with minimal effort from the user. However, despite AutoML showing great promise for computer vision tasks, to the best of our knowledge, no study has used AutoML for image-based plant phenotyping. To address this gap in knowledge, we examined the application of AutoML for image-based plant phenotyping using wheat lodging assessment with unmanned aerial vehicle (UAV) imagery as an example. The performance of an open-source AutoML framework, AutoKeras, in image classification and regression tasks was compared to transfer learning using modern convolutional neural network (CNN) architectures. For image classification, which classified plot images as lodged or non-lodged, transfer learning with Xception and DenseNet-201 achieved the best classification accuracy of 93.2%, whereas AutoKeras had a 92.4% accuracy. For image regression, which predicted lodging scores from plot images, transfer learning with DenseNet-201 had the best performance (<i>R</i><sup>2</sup> = 0.8303, root mean-squared error (<i>RMSE</i>) = 9.55, mean absolute error (<i>MAE</i>) = 7.03, mean absolute percentage error (<i>MAPE</i>) = 12.54%), followed closely by AutoKeras (<i>R</i><sup>2</sup> = 0.8273, <i>RMSE</i> = 10.65, <i>MAE</i> = 8.24, <i>MAPE</i> = 13.87%). In both tasks, AutoKeras models had up to 40-fold faster inference times compared to the pretrained CNNs. AutoML has significant potential to enhance plant phenotyping capabilities applicable in crop breeding and precision agriculture. |
first_indexed | 2024-03-09T00:32:05Z |
format | Article |
id | doaj.art-c88b2daa8d084e92abd49dae401b9403 |
institution | Directory Open Access Journal |
issn | 2072-4292 |
language | English |
last_indexed | 2024-03-09T00:32:05Z |
publishDate | 2021-02-01 |
publisher | MDPI AG |
record_format | Article |
series | Remote Sensing |
spelling | doaj.art-c88b2daa8d084e92abd49dae401b94032023-12-11T18:26:53ZengMDPI AGRemote Sensing2072-42922021-02-0113585810.3390/rs13050858Automated Machine Learning for High-Throughput Image-Based Plant PhenotypingJoshua C.O. Koh0German Spangenberg1Surya Kant2Agriculture Victoria, Grains Innovation Park, 110 Natimuk Rd, Horsham, VIC 3400, AustraliaAgriculture Victoria, AgriBio, Centre for AgriBioscience, 5 Ring Road, Bundoora, VIC 3083, AustraliaAgriculture Victoria, Grains Innovation Park, 110 Natimuk Rd, Horsham, VIC 3400, AustraliaAutomated machine learning (AutoML) has been heralded as the next wave in artificial intelligence with its promise to deliver high-performance end-to-end machine learning pipelines with minimal effort from the user. However, despite AutoML showing great promise for computer vision tasks, to the best of our knowledge, no study has used AutoML for image-based plant phenotyping. To address this gap in knowledge, we examined the application of AutoML for image-based plant phenotyping using wheat lodging assessment with unmanned aerial vehicle (UAV) imagery as an example. The performance of an open-source AutoML framework, AutoKeras, in image classification and regression tasks was compared to transfer learning using modern convolutional neural network (CNN) architectures. For image classification, which classified plot images as lodged or non-lodged, transfer learning with Xception and DenseNet-201 achieved the best classification accuracy of 93.2%, whereas AutoKeras had a 92.4% accuracy. For image regression, which predicted lodging scores from plot images, transfer learning with DenseNet-201 had the best performance (<i>R</i><sup>2</sup> = 0.8303, root mean-squared error (<i>RMSE</i>) = 9.55, mean absolute error (<i>MAE</i>) = 7.03, mean absolute percentage error (<i>MAPE</i>) = 12.54%), followed closely by AutoKeras (<i>R</i><sup>2</sup> = 0.8273, <i>RMSE</i> = 10.65, <i>MAE</i> = 8.24, <i>MAPE</i> = 13.87%). In both tasks, AutoKeras models had up to 40-fold faster inference times compared to the pretrained CNNs. AutoML has significant potential to enhance plant phenotyping capabilities applicable in crop breeding and precision agriculture.https://www.mdpi.com/2072-4292/13/5/858automated machine learningneural architecture searchhigh-throughput plant phenotypingwheat lodging assessmentunmanned aerial vehicle |
spellingShingle | Joshua C.O. Koh German Spangenberg Surya Kant Automated Machine Learning for High-Throughput Image-Based Plant Phenotyping Remote Sensing automated machine learning neural architecture search high-throughput plant phenotyping wheat lodging assessment unmanned aerial vehicle |
title | Automated Machine Learning for High-Throughput Image-Based Plant Phenotyping |
title_full | Automated Machine Learning for High-Throughput Image-Based Plant Phenotyping |
title_fullStr | Automated Machine Learning for High-Throughput Image-Based Plant Phenotyping |
title_full_unstemmed | Automated Machine Learning for High-Throughput Image-Based Plant Phenotyping |
title_short | Automated Machine Learning for High-Throughput Image-Based Plant Phenotyping |
title_sort | automated machine learning for high throughput image based plant phenotyping |
topic | automated machine learning neural architecture search high-throughput plant phenotyping wheat lodging assessment unmanned aerial vehicle |
url | https://www.mdpi.com/2072-4292/13/5/858 |
work_keys_str_mv | AT joshuacokoh automatedmachinelearningforhighthroughputimagebasedplantphenotyping AT germanspangenberg automatedmachinelearningforhighthroughputimagebasedplantphenotyping AT suryakant automatedmachinelearningforhighthroughputimagebasedplantphenotyping |