Sugar Cane Detection

Project overview

Sugar cane field imagery was annotated by boxing cane plants and stalks so that a model could support stand counting and crop monitoring. The dataset was prepared as training data for a model that detects and counts sugar cane for stand and growth monitoring.

Work performed

• Reviewed and organised the source imagery before annotation.
• Identified and localised every visible sugar cane plant or stalk in each image.
• Drew tight, accurate bounding boxes around each target object.
• Boxed each cane plant and stalk individually, including within dense stands.
• Applied class labelling consistently across the whole dataset.
• Performed quality checks and corrected inaccurate or missing annotations.
• Ensured the annotations were suitable for computer vision model training.

What made it hard

• Very dense, overlapping stalks.
• Thin, elongated shapes that were hard to box tightly.
• Cluttered field backgrounds.

Result

A sugar cane detection dataset was completed and prepared to support stand counting and crop monitoring.

Final deliverable

Accurately annotated imagery of sugar cane plants and stalks was delivered, with consistent class labelling and annotation quality throughout the dataset. The final output was prepared to the project requirements and was ready for AI model training and development.