Syngenta Weed Control, Lapley Fields

Project overview

The Syngenta Weed Control project involved annotating agricultural field images from Lapley Fields to identify and segment two major weed classes precisely: Black Grass and Rye Grass. The dataset was prepared to support computer vision model training for weed detection, segmentation and agricultural weed-control applications.

Work performed

• Reviewed the agricultural field images from Lapley Fields.
• Identified Black Grass and Rye Grass plants throughout the imagery.
• Created precise polygon segmentation masks around individual weeds.
• Followed the visible boundaries of leaves and plant structures carefully.
• Differentiated Black Grass from Rye Grass and from surrounding crops and vegetation.
• Handled overlapping and partially visible plants.
• Applied class labelling and segmentation quality consistently across the dataset.
• Reviewed and corrected inaccurate or incomplete annotations.
• Prepared the dataset for weed-control and computer vision model training.

What made it hard

• Differentiating Black Grass from Rye Grass, since the two can look very similar.
• Creating precise segmentation boundaries in dense field vegetation.
• Overlapping leaves and plants complicated polygon placement.
• Segmenting small and partially visible weeds.
• Distinguishing weeds from crops and surrounding vegetation.
• Keeping polygon boundaries consistent across different field conditions and lighting.

Result

A high-quality annotated dataset was prepared for Black Grass and Rye Grass segmentation, supporting computer vision applications for weed identification, weed control and agricultural field monitoring.

Final deliverable

The final dataset contained accurate and consistent polygon segmentation masks for Black Grass and Rye Grass, making it suitable for computer vision model training and agricultural weed-control applications.