A Weakly Supervised Deep Learning Framework for Sorghum Head Detection and Counting
The yield of cereal crops such as sorghum (Sorghum bicolor L. Moench) depends on the distribution of crop-heads in varying branching arrangements. Therefore, counting the head number per unit area is critical for plant breeders to correlate with the genotypic variation in a specific breeding field....
Main Authors: | , , , , , , , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
American Association for the Advancement of Science
2019-01-01
|
Series: | Plant Phenomics |
Online Access: | http://dx.doi.org/10.34133/2019/1525874 |
id |
doaj-1540a75a78bc4f408210dd5791e583a4 |
---|---|
record_format |
Article |
spelling |
doaj-1540a75a78bc4f408210dd5791e583a42020-11-25T00:27:50ZengAmerican Association for the Advancement of SciencePlant Phenomics2643-65152019-01-01201910.34133/2019/1525874A Weakly Supervised Deep Learning Framework for Sorghum Head Detection and CountingSambuddha Ghosal0Sambuddha Ghosal1Bangyou Zheng2Scott C. Chapman3Scott C. Chapman4Andries B. Potgieter5David R. Jordan6Xuemin Wang7Asheesh K. Singh8Arti Singh9Masayuki Hirafuji10Seishi Ninomiya11Baskar Ganapathysubramanian12Soumik Sarkar13Wei Guo14Department of Mechanical Engineering,Iowa State University,Ames,IA,USADepartment of Computer Science,Iowa State University,Ames,IA,USACSIRO Agriculture and Food,St. Lucia,QLD,AustraliaCSIRO Agriculture and Food,St. Lucia,QLD,AustraliaSchool of Agriculture and Food Sciences,The University of Queensland,Gatton,QLD 4343,AustraliaQueensland Alliance for Agriculture and Food Innovation (QAAFI),The University of Queensland,Gatton,QLD,AustraliaQueensland Alliance for Agriculture and Food Innovation (QAAFI),The University of Queensland,Warwick,QLD,AustraliaQueensland Alliance for Agriculture and Food Innovation (QAAFI),The University of Queensland,Warwick,QLD,AustraliaDepartment of Agronomy,Iowa State University,Ames,IA,USADepartment of Agronomy,Iowa State University,Ames,IA,USAInternational Field Phenomics Research Laboratory,Institute for Sustainable Agro-Ecosystem Services,Graduate School of Agricultural and Life Sciences,The University of Tokyo, Tokyo,JapanInternational Field Phenomics Research Laboratory,Institute for Sustainable Agro-Ecosystem Services,Graduate School of Agricultural and Life Sciences,The University of Tokyo, Tokyo,JapanDepartment of Mechanical Engineering,Iowa State University,Ames,IA,USADepartment of Mechanical Engineering,Iowa State University,Ames,IA,USAInternational Field Phenomics Research Laboratory,Institute for Sustainable Agro-Ecosystem Services,Graduate School of Agricultural and Life Sciences,The University of Tokyo, Tokyo,JapanThe yield of cereal crops such as sorghum (Sorghum bicolor L. Moench) depends on the distribution of crop-heads in varying branching arrangements. Therefore, counting the head number per unit area is critical for plant breeders to correlate with the genotypic variation in a specific breeding field. However, measuring such phenotypic traits manually is an extremely labor-intensive process and suffers from low efficiency and human errors. Moreover, the process is almost infeasible for large-scale breeding plantations or experiments. Machine learning-based approaches like deep convolutional neural network (CNN) based object detectors are promising tools for efficient object detection and counting. However, a significant limitation of such deep learning-based approaches is that they typically require a massive amount of hand-labeled images for training, which is still a tedious process. Here, we propose an active learning inspired weakly supervised deep learning framework for sorghum head detection and counting from UAV-based images. We demonstrate that it is possible to significantly reduce human labeling effort without compromising final model performance (R2 between human count and machine count is 0.88) by using a semitrained CNN model (i.e., trained with limited labeled data) to perform synthetic annotation. In addition, we also visualize key features that the network learns. This improves trustworthiness by enabling users to better understand and trust the decisions that the trained deep learning model makes.http://dx.doi.org/10.34133/2019/1525874 |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Sambuddha Ghosal Sambuddha Ghosal Bangyou Zheng Scott C. Chapman Scott C. Chapman Andries B. Potgieter David R. Jordan Xuemin Wang Asheesh K. Singh Arti Singh Masayuki Hirafuji Seishi Ninomiya Baskar Ganapathysubramanian Soumik Sarkar Wei Guo |
spellingShingle |
Sambuddha Ghosal Sambuddha Ghosal Bangyou Zheng Scott C. Chapman Scott C. Chapman Andries B. Potgieter David R. Jordan Xuemin Wang Asheesh K. Singh Arti Singh Masayuki Hirafuji Seishi Ninomiya Baskar Ganapathysubramanian Soumik Sarkar Wei Guo A Weakly Supervised Deep Learning Framework for Sorghum Head Detection and Counting Plant Phenomics |
author_facet |
Sambuddha Ghosal Sambuddha Ghosal Bangyou Zheng Scott C. Chapman Scott C. Chapman Andries B. Potgieter David R. Jordan Xuemin Wang Asheesh K. Singh Arti Singh Masayuki Hirafuji Seishi Ninomiya Baskar Ganapathysubramanian Soumik Sarkar Wei Guo |
author_sort |
Sambuddha Ghosal |
title |
A Weakly Supervised Deep Learning Framework for Sorghum Head Detection and Counting |
title_short |
A Weakly Supervised Deep Learning Framework for Sorghum Head Detection and Counting |
title_full |
A Weakly Supervised Deep Learning Framework for Sorghum Head Detection and Counting |
title_fullStr |
A Weakly Supervised Deep Learning Framework for Sorghum Head Detection and Counting |
title_full_unstemmed |
A Weakly Supervised Deep Learning Framework for Sorghum Head Detection and Counting |
title_sort |
weakly supervised deep learning framework for sorghum head detection and counting |
publisher |
American Association for the Advancement of Science |
series |
Plant Phenomics |
issn |
2643-6515 |
publishDate |
2019-01-01 |
description |
The yield of cereal crops such as sorghum (Sorghum bicolor L. Moench) depends on the distribution of crop-heads in varying branching arrangements. Therefore, counting the head number per unit area is critical for plant breeders to correlate with the genotypic variation in a specific breeding field. However, measuring such phenotypic traits manually is an extremely labor-intensive process and suffers from low efficiency and human errors. Moreover, the process is almost infeasible for large-scale breeding plantations or experiments. Machine learning-based approaches like deep convolutional neural network (CNN) based object detectors are promising tools for efficient object detection and counting. However, a significant limitation of such deep learning-based approaches is that they typically require a massive amount of hand-labeled images for training, which is still a tedious process. Here, we propose an active learning inspired weakly supervised deep learning framework for sorghum head detection and counting from UAV-based images. We demonstrate that it is possible to significantly reduce human labeling effort without compromising final model performance (R2 between human count and machine count is 0.88) by using a semitrained CNN model (i.e., trained with limited labeled data) to perform synthetic annotation. In addition, we also visualize key features that the network learns. This improves trustworthiness by enabling users to better understand and trust the decisions that the trained deep learning model makes. |
url |
http://dx.doi.org/10.34133/2019/1525874 |
work_keys_str_mv |
AT sambuddhaghosal aweaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT sambuddhaghosal aweaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT bangyouzheng aweaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT scottcchapman aweaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT scottcchapman aweaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT andriesbpotgieter aweaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT davidrjordan aweaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT xueminwang aweaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT asheeshksingh aweaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT artisingh aweaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT masayukihirafuji aweaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT seishininomiya aweaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT baskarganapathysubramanian aweaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT soumiksarkar aweaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT weiguo aweaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT sambuddhaghosal weaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT sambuddhaghosal weaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT bangyouzheng weaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT scottcchapman weaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT scottcchapman weaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT andriesbpotgieter weaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT davidrjordan weaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT xueminwang weaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT asheeshksingh weaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT artisingh weaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT masayukihirafuji weaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT seishininomiya weaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT baskarganapathysubramanian weaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT soumiksarkar weaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting AT weiguo weaklysuperviseddeeplearningframeworkforsorghumheaddetectionandcounting |
_version_ |
1716166210688647168 |