Corn Detection & Growth Stage Classification

Project overview

Agricultural field images were annotated for a computer vision dataset focused on detecting and classifying corn plants and weeds. The work involved identifying corn plants at different growth stages and distinguishing them from weeds and visually confusing plants.

Work performed

• Reviewed and organised the agricultural field image dataset before annotation.
• Identified corn plants and weeds throughout the field images.
• Corn plants were classified according to their size and growth stage.
• Created accurate bounding boxes around individual targets.
• Larger plants were labelled as Large Corn and smaller plants as Small Corn.
• Weeds were identified and labelled using the Weed class.
• Applied the Confusing class to plants or objects that were difficult to distinguish with confidence during annotation.
• Differentiated corn plants carefully from surrounding weeds and vegetation.
• Performed quality checks and corrected inaccurate, missing, or poorly positioned annotations.
• Maintained consistent labelling standards throughout the dataset.

What made it hard

• Differentiating corn plants from visually similar weeds.
• Small corn plants were hard to separate from surrounding weeds during early growth stages.
• Distinguishing Large Corn from Small Corn consistently on the basis of plant size and growth stage.
• Handling overlapping and closely positioned plants.
• Confusing cases arose where plant characteristics were unclear.
• Complex field backgrounds contained soil, vegetation, and crop residue.
• Varying lighting conditions and image quality affected plant visibility.
• Accurate bounding boxes and consistent class labels required careful attention to detail.

Result

A high-quality corn detection dataset was completed with accurate and consistent annotations for the Weed, Large Corn, Small Corn, and Confusing classes. The dataset was prepared for training a computer vision model capable of detecting corn plants at different growth stages while distinguishing them from weeds and ambiguous cases.

Final deliverable

Accurate and consistently labelled bounding box annotations were delivered for the Weed, Large Corn, Small Corn, and Confusing classes throughout the dataset. The final output was prepared according to project requirements and was ready for AI model training and automated corn detection, growth stage identification, and weed monitoring applications.