High-throughput soybean seeds phenotyping with convolutional neural networks and transfer learning
Abstract Background Effective soybean seed phenotyping demands large-scale accurate quantities of morphological parameters. The traditional manual acquisition of soybean seed morphological phenotype information is error-prone, and time-consuming, which is not feasible for large-scale collection. The...
Main Authors: | , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
BMC
2021-05-01
|
Series: | Plant Methods |
Subjects: | |
Online Access: | https://doi.org/10.1186/s13007-021-00749-y |
id |
doaj-3c05c3aa99cf41e9bf6c51d67da4f297 |
---|---|
record_format |
Article |
spelling |
doaj-3c05c3aa99cf41e9bf6c51d67da4f2972021-05-09T11:11:24ZengBMCPlant Methods1746-48112021-05-0117111710.1186/s13007-021-00749-yHigh-throughput soybean seeds phenotyping with convolutional neural networks and transfer learningSi Yang0Lihua Zheng1Peng He2Tingting Wu3Shi Sun4Minjuan Wang5College of Information and Electrical Engineering, China Agricultural UniversityCollege of Information and Electrical Engineering, China Agricultural UniversityCollege of Information Engineering, Northwest A&F UniversityInstitute of Crop Sciences, Chinese Academy of Agricultural SciencesInstitute of Crop Sciences, Chinese Academy of Agricultural SciencesCollege of Information and Electrical Engineering, China Agricultural UniversityAbstract Background Effective soybean seed phenotyping demands large-scale accurate quantities of morphological parameters. The traditional manual acquisition of soybean seed morphological phenotype information is error-prone, and time-consuming, which is not feasible for large-scale collection. The segmentation of individual soybean seed is the prerequisite step for obtaining phenotypic traits such as seed length and seed width. Nevertheless, traditional image-based methods for obtaining high-throughput soybean seed phenotype are not robust and practical. Although deep learning-based algorithms can achieve accurate training and strong generalization capabilities, it requires a large amount of ground truth data which is often the limitation step. Results We showed a novel synthetic image generation and augmentation method based on domain randomization. We synthesized a plenty of labeled image dataset automatedly by our method to train instance segmentation network for high throughput soybean seeds segmentation. It can pronouncedly decrease the cost of manual annotation and facilitate the preparation of training dataset. And the convolutional neural network can be purely trained by our synthetic image dataset to achieve a good performance. In the process of training Mask R-CNN, we proposed a transfer learning method which can reduce the computing costs significantly by finetuning the pre-trained model weights. We demonstrated the robustness and generalization ability of our method by analyzing the result of synthetic test datasets with different resolution and the real-world soybean seeds test dataset. Conclusion The experimental results show that the proposed method realized the effective segmentation of individual soybean seed and the efficient calculation of the morphological parameters of each seed and it is practical to use this approach for high-throughput objects instance segmentation and high-throughput seeds phenotyping.https://doi.org/10.1186/s13007-021-00749-ySeed phenotypingHigh throughputInstance segmentationDeep learningMask R-CNN |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Si Yang Lihua Zheng Peng He Tingting Wu Shi Sun Minjuan Wang |
spellingShingle |
Si Yang Lihua Zheng Peng He Tingting Wu Shi Sun Minjuan Wang High-throughput soybean seeds phenotyping with convolutional neural networks and transfer learning Plant Methods Seed phenotyping High throughput Instance segmentation Deep learning Mask R-CNN |
author_facet |
Si Yang Lihua Zheng Peng He Tingting Wu Shi Sun Minjuan Wang |
author_sort |
Si Yang |
title |
High-throughput soybean seeds phenotyping with convolutional neural networks and transfer learning |
title_short |
High-throughput soybean seeds phenotyping with convolutional neural networks and transfer learning |
title_full |
High-throughput soybean seeds phenotyping with convolutional neural networks and transfer learning |
title_fullStr |
High-throughput soybean seeds phenotyping with convolutional neural networks and transfer learning |
title_full_unstemmed |
High-throughput soybean seeds phenotyping with convolutional neural networks and transfer learning |
title_sort |
high-throughput soybean seeds phenotyping with convolutional neural networks and transfer learning |
publisher |
BMC |
series |
Plant Methods |
issn |
1746-4811 |
publishDate |
2021-05-01 |
description |
Abstract Background Effective soybean seed phenotyping demands large-scale accurate quantities of morphological parameters. The traditional manual acquisition of soybean seed morphological phenotype information is error-prone, and time-consuming, which is not feasible for large-scale collection. The segmentation of individual soybean seed is the prerequisite step for obtaining phenotypic traits such as seed length and seed width. Nevertheless, traditional image-based methods for obtaining high-throughput soybean seed phenotype are not robust and practical. Although deep learning-based algorithms can achieve accurate training and strong generalization capabilities, it requires a large amount of ground truth data which is often the limitation step. Results We showed a novel synthetic image generation and augmentation method based on domain randomization. We synthesized a plenty of labeled image dataset automatedly by our method to train instance segmentation network for high throughput soybean seeds segmentation. It can pronouncedly decrease the cost of manual annotation and facilitate the preparation of training dataset. And the convolutional neural network can be purely trained by our synthetic image dataset to achieve a good performance. In the process of training Mask R-CNN, we proposed a transfer learning method which can reduce the computing costs significantly by finetuning the pre-trained model weights. We demonstrated the robustness and generalization ability of our method by analyzing the result of synthetic test datasets with different resolution and the real-world soybean seeds test dataset. Conclusion The experimental results show that the proposed method realized the effective segmentation of individual soybean seed and the efficient calculation of the morphological parameters of each seed and it is practical to use this approach for high-throughput objects instance segmentation and high-throughput seeds phenotyping. |
topic |
Seed phenotyping High throughput Instance segmentation Deep learning Mask R-CNN |
url |
https://doi.org/10.1186/s13007-021-00749-y |
work_keys_str_mv |
AT siyang highthroughputsoybeanseedsphenotypingwithconvolutionalneuralnetworksandtransferlearning AT lihuazheng highthroughputsoybeanseedsphenotypingwithconvolutionalneuralnetworksandtransferlearning AT penghe highthroughputsoybeanseedsphenotypingwithconvolutionalneuralnetworksandtransferlearning AT tingtingwu highthroughputsoybeanseedsphenotypingwithconvolutionalneuralnetworksandtransferlearning AT shisun highthroughputsoybeanseedsphenotypingwithconvolutionalneuralnetworksandtransferlearning AT minjuanwang highthroughputsoybeanseedsphenotypingwithconvolutionalneuralnetworksandtransferlearning |
_version_ |
1721454581931573248 |