Black Grass Segmentation

Project overview

The Vissenbjreg_3 Weed Control project involved segmenting black grass precisely within agricultural field images. The main objective was to identify black grass plants growing among crops and prepare accurate segmentation data for weed control and computer vision model training.

Work performed

• Reviewed agricultural field images and identified black grass plants.
• Created precise polygon segmentation masks around each black grass plant.
• Followed the visible boundaries of individual plants and leaves carefully.
• Differentiated black grass from crops, soil and other surrounding vegetation.
• Overlapping and partially visible plants were handled case by case.
• Maintained consistent segmentation quality throughout the dataset.
• Reviewed and corrected inaccurate or incomplete masks.
• Prepared the dataset for weed control and segmentation model training.

What made it hard

• Distinguishing black grass from similar-looking grasses and crops.
• Creating precise boundaries in dense agricultural fields.
• Overlapping leaves and plants complicated mask placement.
• Small and partially visible black grass plants were difficult to segment.
• Complex field backgrounds and varying lighting conditions added difficulty.
• Keeping polygon boundaries accurate and consistent across the dataset.

Result

A high-quality annotated dataset was prepared for black grass segmentation, supporting computer vision applications for weed detection, weed control and agricultural field monitoring.

Final deliverable

The final dataset contained accurate and consistent segmentation masks for the Black Grass class and was prepared for computer vision model training and agricultural weed control applications.