Developing AI to bridge lab and field plant research
Lead Research Organisation:
National Institute of Agricultural Botany
Department Name: Centre for Research
Abstract
The imminent challenges of climate change, growing food demand, and fertiliser shortage have brought enormous threats to global food security. As one of the most consumed grains in the world, wheat (Triticum aestivum L.) and its sustainable production are paramount to ensure food supply. The early parts of wheat developmental phase, i.e. germination (growth stage, GS 00-09) and seedling development (GS 10-19), are critical as a poor-quality establishment in the field can translate into: (1) a reduced plant density and thus a lower yield production, or (2) decreased competitiveness of crops against weeds and the development of diseases. In fact, better seed quality and vigour often lead to improved crop performance and health, ensuring crop sustainability under complex field conditions. Due to the importance of germination and seedling, the foundation phase in wheat underpins modern breeding, cultivation, agronomy, and even smart agricultural activities.
Still, some lab-based research discoveries in seed vigour could vanish when they were moved to the field, which might be caused by genetics, environmental factors, agronomic management, and other in-field matters. AI-powered solutions could help the selection of the most predictive features from lab-based seed quality and vigour characteristics through a range of supervised or unsupervised algorithms, followed by the connection between lab-based features with in-field seedling performance. The above presents opportunities that AI could help bridge the gap between lab-based and field-based plant research:
To identify spectral signatures (i.e. reflectance measured using single or multiple wavelengths to signify internal components of seeds) from multispectral seed imaging to assess seed quality.
To establish AI-powered tracking methods to quantify radicle emergence and seedling establishment to quantify and classify seed vigour.
To build a feature selection approach to select the most relevant features identified through lab-based experiments to predict and verify seedling development under field conditions.
Both NIAB UK (led by Prof Ji Zhou) and Université d'Angers (University of Angers France, same below; led by Prof David Rousseau) have been developing AI solutions in areas such as multi-spectral seed imaging to assess seed quality, seed germination tests for seed vigour assessment, and drone-based phenotyping to study crop early establishment. This "Partner with international researchers on AI for Bioscience" call will provide a unique opportunity for both sides to work together, learning from counterpart's AI solutions, jointly optimising the existing toolkits, and more importantly, seeking and developing AI-powered solutions to connect lab-based seed quality and vigour assessment with in-field seedling establishment using wheat as a model plant. We trust the proposed project will be a valuable case study that demonstrates how to combine AI and domain knowledge to bridge the gap between laboratory and field-based plant research.
Still, some lab-based research discoveries in seed vigour could vanish when they were moved to the field, which might be caused by genetics, environmental factors, agronomic management, and other in-field matters. AI-powered solutions could help the selection of the most predictive features from lab-based seed quality and vigour characteristics through a range of supervised or unsupervised algorithms, followed by the connection between lab-based features with in-field seedling performance. The above presents opportunities that AI could help bridge the gap between lab-based and field-based plant research:
To identify spectral signatures (i.e. reflectance measured using single or multiple wavelengths to signify internal components of seeds) from multispectral seed imaging to assess seed quality.
To establish AI-powered tracking methods to quantify radicle emergence and seedling establishment to quantify and classify seed vigour.
To build a feature selection approach to select the most relevant features identified through lab-based experiments to predict and verify seedling development under field conditions.
Both NIAB UK (led by Prof Ji Zhou) and Université d'Angers (University of Angers France, same below; led by Prof David Rousseau) have been developing AI solutions in areas such as multi-spectral seed imaging to assess seed quality, seed germination tests for seed vigour assessment, and drone-based phenotyping to study crop early establishment. This "Partner with international researchers on AI for Bioscience" call will provide a unique opportunity for both sides to work together, learning from counterpart's AI solutions, jointly optimising the existing toolkits, and more importantly, seeking and developing AI-powered solutions to connect lab-based seed quality and vigour assessment with in-field seedling establishment using wheat as a model plant. We trust the proposed project will be a valuable case study that demonstrates how to combine AI and domain knowledge to bridge the gap between laboratory and field-based plant research.
People |
ORCID iD |
| Ji Zhou (Principal Investigator) |
 http://orcid.org/0000-0002-5752-5524
|
Publications
Percival-Alwyn L
(2024)
UKCropDiversity-HPC: A collaborative high-performance computing resource approach for sustainable agriculture and biodiversity conservation
in PLANTS, PEOPLE, PLANET
| Description | We found that the multispectral can lead to more precise genetic mapping studies. For example, from red to red edge and near infrared channels can be soundly used to identify different spectral reflectances of genotypes in the Wheat MAGIC Diversity Panel developed by NIAB. We identified several novel and published genetic loci, some of which (e.g. Tamyb10-B1, reported to change grain colour from red to white, showing higher grain dormancy) took plant researcher 7 years to identify. The GWAS result was only 5kb away from the cloned gene, suggesting the breeding value of this approach. In 2024, we also established SeedGerm-VIG, an automated and comprehensive pipeline developed for assessing seed vigour in wheat. Building on the SeedGerm platform, we integrated optimised deep learning models (i.e. YOLOv8s-Germ and U-Net) into seed-level analysis to identify key germination phases and quantify seed, root, and seedling traits. Moreover, using time series directed graph, not only did we track root tips to measure root emergence during the germination procedure, but we also established a new approach to examine speed and uniformity of seed germination. These resulted in the establishment of a seed vigour scoring matrix, through which 21 commercial genotypes' vigour scores were evaluated based on their dynamic features during protrusion, radicle emergence, and chloroplast biogenesis. Finally, we successfully applied the SeedGerm-VIG pipeline to assess seed vigour for other cereal crops such as rice and barley with high-accuracy analysis results. We believe that our work demonstrates a valuable advance to enable the broader plant and crop research community to quantify seed vigour and vigour-related features in an automated and reproducible manner, facilitating effective plant selection and relevant seed science research for crop improvement under global climate change. |
| Exploitation Route | After this year's (2025) review and improvement, I think we could soundly provide the AI models, training datasets, and software platform to the community to utilise, so that they can be utilised and extended by other plant and crop researchers, as well as breeders. |
| Sectors | Agriculture Food and Drink Digital/Communication/Information Technologies (including Software) |
| URL | https://github.com/The-Zhou-Lab/SeedGerm-VIG |
| Description | ISTA seed vigour testing as the SeedGerm-VIG can help ISTA develop and validate vigour tests following the ISTA International Rules for Seed Testing, which can encourage the use of vigour tests by seed laboratories worldwide more effectively. We also hope the SeedGerm-VIG can be used to provide information and training to ensure that vigour tests are carried out to achieve reproducibility and scalability in laboratories, in 80+ regions and countries around the world. |
| First Year Of Impact | 2025 |
| Sector | Agriculture, Food and Drink,Digital/Communication/Information Technologies (including Software) |
| Impact Types | Economic Policy & public services |
| Description | AI in Digital Agriculture and Sustainable Agriculture |
| Geographic Reach | Multiple continents/international |
| Policy Influence Type | Influenced training of practitioners or researchers |
| Impact | One CGIAR is a global research partnership for a food secure future, dedicated to transforming food, land and water systems in a climate crisis through a $1 billion annual research portfolio. The partnership unites 15 international organisations and has 10,000 staff working across 89 different countries, producing science that has brought benefits to hundreds of millions of people around the world. One CGIAR supports innovative solutions to improve food security, increase biodiversity, stimulate economic growth and strengthen resilience of farming systems. AI-powered toolkit was developed to facilitate the accurate measurement of adoption rates of CGIAR varietal technologies and the prediction of yield to better measure genetic gain in farmer fields. And this is where the AI/Data Sciences team at NIAB comes in. Currently, NIAB has pioneered AI-models to assess key yield components in wheat and are developing AI-based image recognition algorithms and open-source software. With One CGIAR the ambition is to integrate this into VarScout, a digital ecosystem designed to record, store, and visualise crop varietal data on a local, national and international scale. |
| URL | https://www.niab.com/news-views/blogs/using-artificial-intelligence-variety-identification |
| Description | Provided training to PhD students at the University of Angers |
| Geographic Reach | Europe |
| Policy Influence Type | Influenced training of practitioners or researchers |
| Impact | Students were happy to see how to speed up their AI modelling using optimised learning architecture instead of using GPU-clusters (i.e. hardware) only, as well as how to build high-quality training datasets. Students trailed them at the lab and emailed Prof Zhou if they had problems. |
| URL | https://www.univ-angers.fr/fr/formations/actualites/actus-2024/concours-rousseau-l-equipe-de-l-ua-s-... |
| Title | GSP-AI model |
| Description | We uploaded the latest version of GSP-AI model, iPython, testing data, and other supporting datasets (see the Assets section below) that could be used to reproduce the results presented in the article titled "GSP-AI: An AI-Powered Platform for Identifying Key Growth Stages and the Vegetative-to-Reproductive Transition in Wheat Using Trilateral Drone Imagery and Meteorological Data" (DOI: 10.34133/plantphenomics.0255; https://spj.science.org/doi/10.34133/plantphenomics.0255). Other data and user guides are openly available on request. The latest AirMeasurer platform can be downloaded via https://github.com/The-Zhou-Lab/UAV-AirMeasurer/releases). |
| Type Of Material | Improvements to research infrastructure |
| Year Produced | 2024 |
| Provided To Others? | Yes |
| Impact | Testing datasets, software, wheat growth stage prediction modelling, and manually scored flowering date provided in this release are available for academic use only. Other data and user guides are openly available on request. When using the Zhou lab's work, please do cite our paper "GSP-AI: An AI-powered platform for identifying key growth stages and the vegetative-to-reproductive transition in wheat based on trilateral and multi-seasonal drone phenotyping", as well as the AirMeasurer platform (DOI: 10.1111/nph.18314). |
| URL | https://github.com/The-Zhou-Lab/GSP-AI/releases |
| Title | New drone based method has been developed to study seedling with drone-collected images |
| Description | OMEX supplied four UK recommended varieties with different yield production (e.g. Dart, Flemming, and Aurelia), with or without JA treatment. In the pre-germinating assessment, we used the Videometer system and its 19 probes to produce wavelengths from 365 nm to 970 nm plus fluorescence to image dry and imbibed seeds. These spectral traits were then used as proxies to correlate with seed quality (e.g. uniformity, etc.) and thus the classification of seed quality for the four UK varieties with or without JA treatment. This was then used as benchmark data for the field-based seedling establishment analysis. |
| Type Of Material | Physiological assessment or outcome measure |
| Year Produced | 2024 |
| Provided To Others? | No |
| Impact | Established seedling phenotyping in the field, with seedling trait analysis generated for different rapeseed varieties, with or without JA. |
| Title | The SeedGerm-VIG platform |
| Description | The latest version of SeedGerm-VIG pipeline, source code, testing data, and other supporting datasets (see the Assets section via https://github.com/The-Zhou-Lab/SeedGerm-VIG/releases/tag/v1.0). |
| Type Of Material | Improvements to research infrastructure |
| Year Produced | 2025 |
| Provided To Others? | Yes |
| Impact | SeedGerm-VIG can be used to reproduce the results presented in the manuscript titled "SeedGerm-VIG: an open and comprehensive methodology to quantify seed vigour in wheat and other cereal crops using deep learning powered dynamic phenotyping". We provided open training datasets and user guides, testing datasets, example germination image series, and models provided in this release are available for academic use only. This work follows the lab's pervious paper "SeedGerm-VIG: an open and comprehensive methodology to quantify seed vigour in wheat and other cereal crops using deep learning powered dynamic phenotyping". |
| URL | https://github.com/The-Zhou-Lab/SeedGerm-VIG/releases/tag/v1.0 |
| Title | Hyperspectral seed imaging analysis |
| Description | To safeguard wheat seed quality and vigour for better crop performance, it is important to assess seed morphological features and internal contents reliably and at a large scale, resulting in the importance of rapid and non-destructive seed analysis and the necessity of advancing analytical methods in this research domain. We have combined multispectral seed imaging, computer vision and automated image processing techniques to address methodological problems in seed phenotyping and seed-based phenotypic analysis in terms of throughput and accuracy. We have developed an automated algorithm that can segment individual seed from hundreds of wheat seeds acquired by multispectral imaging device, through which morphological traits (e.g. seed area, length, width, and roundness) and internal components (e.g. plant pigments, starch, vegetable oil and water content, etc.) can be quantified. We utilised morphological and spectral traits in clustering and principal component analysis, establishing a classification method to differentiate wheat seed varieties. |
| Type Of Material | Computer model/algorithm |
| Year Produced | 2024 |
| Provided To Others? | No |
| Impact | The paper has published in 2024 |
| URL | https://d.wanfangdata.com.cn/periodical/zwslxtx202404013 |
| Title | SeedGerm-VIG |
| Description | The latest version of SeedGerm-VIG pipeline, source code, testing data, and other supporting datasets (see the Assets section via the link below) for reproducing the results presented in the manuscript titled "SeedGerm-VIG: an open and comprehensive methodology to quantify seed vigour in wheat and other cereal crops using deep learning powered dynamic phenotyping". |
| Type Of Material | Computer model/algorithm |
| Year Produced | 2025 |
| Provided To Others? | Yes |
| Impact | We provided resources such as testing datasets, example germination image series, and AI models in the release for academic use only. Other data and user guides are openly available on request. We built the work based on our previous paper "SeedGerm-VIG: an open and comprehensive methodology to quantify seed vigour in wheat and other cereal crops using deep learning powered dynamic phenotyping". |
| URL | https://github.com/The-Zhou-Lab/SeedGerm-VIG/releases/tag/v1.0 |
| Description | Developed a partner with SeedEqual programme at One CGIAR |
| Organisation | International Potato Center |
| Country | Peru |
| Sector | Charity/Non Profit |
| PI Contribution | Digital tools drive for varietal detection under field conditions |
| Collaborator Contribution | Through the CGIAR initiative award, we are contributing to "Combining cloud- and smartphone-based AI solutions for varietal identification of wheat and potato", a project aims to advance the vision-based AI modelling tailored for varietal identification using cutting-edge AI algorithms such as Transformer and Visual Attention. We used some achieved AI models created through BBSRC funding and demonstrated to CGIAR that AI models could be used to assess key yield components in wheat and potato using AI-powered image recognition algorithms, so that new AI toolkits can be integrated into SeedEuqal's VarScout programme (World Bank funding), a digital ecosystem designed to record, store, and visualise crop varietal data on a local, national and international scale for developing countries. Through this partnership, we will help smallholder farmers in developing countries to identify varieties they are growing, which can complement DNA fingerprinting that are very costly and results are only available after months. |
| Impact | We are working on an open-source publication and software (https://github.com/The-Zhou-Lab/GSP-AI), which will be followed with the release of VarScout, one of CGIAR's flagship programmes. |
| Start Year | 2024 |
| Description | INRAe Angers Seed Testing Center |
| Organisation | University of Angers |
| Country | France |
| Sector | Academic/University |
| PI Contribution | We made contribution regarding the multispectral seed imaging can lead to more precise genetic mapping studies. |
| Collaborator Contribution | For example, from red to red edge and near infrared channels can be soundly used to identify different spectral reflectances of genotypes in the Wheat MAGIC Diversity Panel developed by NIAB. We identified several novel and published genetic loci, some of which (e.g. Tamyb10-B1, reported to change grain colour from red to white, showing higher grain dormancy) took plant researcher 7 years to identify. The GWAS result was only 5kb away from the cloned gene, suggesting the breeding value of this approach. We shared our spectral signature algorithm with the INRAe Angers Seed Testing Center and similar experiments are being conducted by the French side, which is likely lead to the identification of molecular markers that INRAe is interested in.... |
| Impact | The work is still on going, this is just the first year of this collaboration. |
| Start Year | 2024 |
| Description | Workshop at the University of Angers |
| Organisation | University of Angers |
| Country | France |
| Sector | Academic/University |
| PI Contribution | The team, Prof Ji Zhou, Dr Rob Jackson, and Dr Arthur Mitchell, have been to Angers, the university and the trial centre, presenting the collaboration, our research in AI and crop improvement, and jointly holding a workshop with INRAe Angers, University of Angers, and several local universities. |
| Collaborator Contribution | Euro 5,000 has been spent by the partner to support joint research activities to study agricultural (wheat) and horticultural corps (e.g. apples and pears). An in-kind contribution has been supplied by the partner (the PhD student and two faculty members' time to work on result analysis and software R&D). |
| Impact | I will add |
| Start Year | 2024 |
| Title | GSP-AI software |
| Description | GSP-AI model, source code, testing data, and other supporting datasets (see the Assets section below) that could be used to reproduce the results presented in the article titled "GSP-AI: An AI-powered platform for identifying key growth stages and the vegetative-to-reproductive transition in wheat based on trilateral and multi-seasonal drone phenotyping". |
| Type Of Technology | Software |
| Year Produced | 2024 |
| Open Source License? | Yes |
| Impact | We first established an open Wheat Growth Stage Prediction (WGSP) dataset, consisting of 70,410 annotated images collected from 54 varieties cultivated in China, 109 in the United Kingdom, and 100 in the United States together with key climatic factors. Then, we built an effective learning architecture based on Res2Net and long short-term memory (LSTM) to learn canopy-level vision features and patterns of climatic changes between 2018 and 2021 growing seasons. Utilizing the model, we achieved an overall accuracy of 91.2% in identifying key GS and an average root mean square error (RMSE) of 5.6 d for forecasting the flowering days compared with manual scoring. We further tested and improved the GSP-AI model with high-resolution smartphone images collected in the 2021/2022 season in China, through which the accuracy of the model was enhanced to 93.4% for GS and RMSE reduced to 4.7 d for the flowering prediction. As a result, we believe that our work demonstrates a valuable advance to inform breeders and growers regarding the timing and duration of key plant growth and development phases at the plot level, facilitating them to conduct more effective crop selection and make agronomic decisions under complicated field conditions for wheat improvement. |
| URL | https://spj.science.org/doi/10.34133/plantphenomics.0255 |
| Title | Hyperspectral imaging for seed quality |
| Description | Established hyperspectral seed imaging and time-series seed germination experiments, with radical and seedling trait analysis generated for different rapeseed varieties, with or without JA. The development is on going... |
| Type Of Technology | New/Improved Technique/Technology |
| Year Produced | 2024 |
| Impact | Time-series seed imaging will demonstrate a new approach for us to assess radical and seedling trait analysis generated for different rapeseed varieties., which can be expanded to other crop species. |
| URL | https://www.geves.fr/research-development/research-activities/seed-quality-testing/ |
| Description | Digital Agriculture For A Sustainable Future |
| Form Of Engagement Activity | Participation in an activity, workshop or similar |
| Part Of Official Scheme? | No |
| Geographic Reach | National |
| Primary Audience | Professional Practitioners |
| Results and Impact | I was invited to present Multi-scale crop phenotyping and AI-powered trait analysis to connect latest technologies with farmers as a keynote speaker. The event is called "Digital Agriculture For A Sustainable Future" which was jointly organised by Hutchinson Limited and Defra in London UK. |
| Year(s) Of Engagement Activity | 2024 |
| URL | https://www.hutchinsons.co.uk/digital-tools-drive-for-a-more-profitable-and-sustainable-farming-futu... |
| Description | Novo Nordisk Foundation presentation |
| Form Of Engagement Activity | Participation in an activity, workshop or similar |
| Part Of Official Scheme? | No |
| Geographic Reach | International |
| Primary Audience | Professional Practitioners |
| Results and Impact | As an invited speaker, I was presented at the Novo Nordisk Foundation workshop at the University of Cambridge UK. The title is "From image to loci: Multi-scale phenotyping and AI-powered trait analysis to enable diversity analysis and genetic mapping of performance traits in wheat". |
| Year(s) Of Engagement Activity | 2025 |