Johnson Grass Detection

Project overview

Agricultural field images were annotated to identify Johnson grass and separate it from other vegetation and objects present in the field. The dataset was prepared for computer vision model training focused on accurate weed detection.

Work performed

• Reviewed agricultural field images and identified Johnson grass plants.
• Accurate bounding box annotations were created around Johnson grass.
• Vegetation and objects outside the Johnson grass class were labelled as Other.
• Johnson grass was differentiated from surrounding crops, weeds and field vegetation.
• Annotation quality remained consistent across the dataset.
• Inaccurate or missing annotations were reviewed and corrected.
• The dataset was prepared for weed detection model training.

What made it hard

• Distinguishing Johnson grass from visually similar vegetation.
• Dense and complex agricultural backgrounds reduced clarity.
• Partially visible or overlapping plants had to be identified.
• Plants appeared at different sizes and growth stages.
• Consistent boundaries and class labels had to be maintained across images.

Result

A high-quality annotated dataset was prepared for Johnson grass detection, suitable for training computer vision models for agricultural weed identification and monitoring.

Final deliverable

The final dataset contained accurate and consistent bounding box annotations for the Johnson_Grass and Other classes, making it suitable for computer vision model training and agricultural weed detection applications.