Broccoli Stand Count

Project overview

Agricultural field images were annotated for a computer vision dataset focused on broccoli stand counting. The work involved identifying individual broccoli plants and creating accurate annotations to provide reliable training data for an AI model designed to detect and count broccoli plants automatically in field environments.

Work performed

• Reviewed and organised the broccoli field image dataset before annotation began.
• Identified individual broccoli plants in each image.
• Created accurate bounding boxes around every visible broccoli plant.
• Maintained consistent labelling using the Broccoli class.
• Partially visible plants were annotated wherever applicable.
• Quality checks were carried out and missed or inaccurate annotations were corrected.
• Annotations were confirmed as suitable for automated plant detection and stand-counting model training.

What made it hard

• Detecting individual broccoli plants in dense field conditions.
• Overlapping and closely positioned plants had to be separated.
• Broccoli plants were easily confused with weeds and surrounding vegetation.
• Partially visible plants near image boundaries still required annotation.
• Accuracy and consistency had to hold across varied field conditions and camera angles.

Result

A high-quality broccoli stand count dataset was completed with accurate annotations for individual plants. The dataset was prepared for training a computer vision model capable of detecting and counting broccoli plants automatically in agricultural fields.

Final deliverable

Accurately annotated broccoli field images were delivered with bounding boxes around individual plants, with consistent labelling and annotation quality maintained throughout the dataset. The final dataset met the project requirements and was ready for AI model training and automated broccoli stand-counting applications.