Multimodal computer vision with convolutional neural networks (CNNs) for crop production

Applicants are invited for a PhD fellowship/scholarship at Graduate School of Technical Sciences, Aarhus University, Denmark, within the Agroecology programme. The position is available from 15 February 2025 or later. You can submit your application via the link under 'how to apply'.

Title
Multimodal computer vision with convolutional neural networks (CNNs) for crop production

Research area and project description
Recent advancements in proximal sensing technologies, such as high-resolution RGB sensors mounted on autonomous robots and farming equipment, have enabled precise detection and classification of crops and weeds in precision agriculture. These systems provide vital data for improving the efficiency and sustainability of farming operations.

This PhD project aims to develop multimodal computer vision systems leveraging hierarchical classification and multilevel spatial mapping to enhance crop and weed detection and classification.

Building on RGB computer-vision data, the PhD research will focus on improving the agricultural knowledge base for crop modeling and optimizing the use of key agricultural inputs, such as nutrients, water, and pesticides. The core objective is to differentiate cultivated crops from non-cultivated species (weeds) and further develop hierarchical classification models to classify non-cultivated species, identifying those that either harm or benefit crop yields and biodiversity.

The core of this PhD project is to develop a comprehensive agricultural intelligence system powered by machine learning, which integrates multiple spatial and temporal mapping levels. By drawing from high-resolution RGB image data and advanced convolutional neural networks (CNNs), the project aims to track and model the dynamics of crops and weeds over time and space. The integration of these mapping techniques will generate actionable insights, allowing for precise identification and classification during the entire growth season. This agricultural intelligence will support the development of smart Integrated Weed Management (IPM) systems, optimizing the use of resources such as water, nutrients, and pesticides, while minimizing environmental impact.

Key tasks will include:

• Developing and optimizing hierarchical classification frameworks to refine non-crop and crop identification, building on state-of-the-art methods in precision agriculture and state-of-the-art object detection models for weed and crop detection.
• Implementing multimodal data processing techniques to improve the accuracy and efficiency of crop and non-crop detection across different growth stages, building on multi-level mapping concepts.
• Utilizing advanced data augmentation strategies to address challenges such as class imbalance and under-utilized data.
• Collaborating with interdisciplinary teams on real-world experiments using robotic platforms and proximal sensing technologies in experimental field settings to validate and refine the developed models.

The Ph.D. candidate must have some experience or interest in working with weed images. It is a further advantage if the Ph.D. candidate has experience in annotating weeds from images. The PhD candidate will be able to develop related research questions based on their interests.

This Ph.D. position offers an opportunity to collaborate inter-disciplinary with other scientists on a common goal to develop sustainable crop production with a high degree of technology.

The Ph.D. is funded by the project ‘System-based Precision Agriculture for Sustainable Crop Production’ and has received grants from the Novo Nordic Foundation.

Project description: For technical reasons, you must upload a project description. Please simply copy the project description above and upload it as a PDF in the application.

Qualifications and specific competences
We seek a highly motivated candidate with an MSc in computer vision, weed science, agronomy, environmental sciences, ecology, or a closely related field. The ideal candidate should have:

• Experience with machine learning models and deep learning frameworks such as CNNs, especially in weed and crop detection.
• Familiarity with hierarchical classification systems and data augmentation techniques, including experience with YOLO models or similar frameworks.
• Strong scientific and technical writing skills with the ability to publish in international journals and present at conferences.
• Enthusiasm for interdisciplinary collaboration and sustainable agricultural practices.

Experience with annotating images of weeds and crops, or working with spatial mapping systems in agriculture, is an advantage.

Place of employment and place of work
The place of employment is Aarhus University, and the place of work is Finlandsgade 22, DK-8200 Aarhus N, Denmark

Contacts
Applicants seeking further information are invited to contact:

• René Gislum, [email protected] (main supervisor)
• Rasmus N. Jørgensen, [email protected] (co-supervisor)

How to apply
Please follow this link to submit your application.

Application deadline is 15 December 2024 at 23:59.

Preferred starting date is 15 February 2025.

For information about application requirements and mandatory attachments, please see our application guide.

Please note:

• Only documents received prior to the application deadline will be evaluated. Thus, documents sent after deadline will not be taken into account.
• The programme committee may request further information or invite the applicant to attend an interview.
• Shortlisting will be used, which means that the evaluation committee only will evaluate the most relevant applications.

Aarhus University’s ambition is to be an attractive and inspiring workplace for all and to foster a culture in which each individual has opportunities to thrive, achieve and develop. We view equality and diversity as assets, and we welcome all applicants. All interested candidates are encouraged to apply, regardless of their personal background.