← Industry Showcase

Agriculture

MacDon Industries Ltd.

Synthetic Data Augmentation and Domain Adaptation for Training Agricultural Semantic Segmentation Models

Showcase video

Showcase presentation — 5 min 56 s

Infographic

Download
MacDon Industries Ltd. infographic

About this showcase

Our research team is developing deep learning (DL) methods to enhance how agricultural machines, like cultivators, detect and navigate crop rows. This work draws on our longstanding interest in applying DL and computer vision to agricultural challenges and builds on prior collaborations with industry. Central to the project is semantic segmentation, a computer vision technique that allows machines to interpret each part of an image and accurately identify where crop rows begin and end. This level of precision is essential for enabling autonomous farm equipment to perform tasks such as weed control and soil management more effectively.

We are working with MacDon Industries Ltd., a leading Canadian agricultural equipment manufacturer, whose goal is to reduce the time and cost of manually labeling training data so they can develop DL algorithms - such as the crop detection system. Together, we are investigating the use of synthetic data - photorealistic images generated by DL models - to augment MacDon's proprietary datasets. This approach promises to accelerate training processes while preserving the quality of DL-based systems. Our work aims to produce scalable, reusable tools and pipelines to support DL development in agriculture and other industrial domains.

The project is being carried out by an interdisciplinary team that includes Drs. Henry and Wang, trainees from the University of Manitoba, and engineers and developers at MacDon. The team brings together expertise in DL, digital agriculture, simulation, and embedded systems. One highly qualified personnel - a recent MSc graduate in computer science - is working full time on this research, with support from an MSc student, along with both academic and industry mentors.

The long-term outcomes will include more capable machine systems, new career opportunities in AI and robotics, and broader economic benefits for Canada's rural and tech sectors. This work is a step forward in making Canadian agriculture more sustainable, efficient, and globally competitive.

At a glance

  • Company MacDon Industries Ltd.
  • SectorAgriculture
  • Presented byChristopher Henry, University of Manitoba

Contact

Christopher Henry

Christopher.Henry@umanitoba.ca