POA Detection

Project overview

The POA Detection project involved annotating agricultural field images to identify POA (Poa) plants. The dataset was prepared for computer vision model training focused on accurate detection and monitoring of POA in agricultural environments.

Work performed

• Reviewed agricultural field images and identified POA plants.
• Accurate bounding box annotations were created around each plant.
• POA was differentiated from surrounding crops, weeds and field vegetation.
• Annotated partially visible plants and plants of differing sizes where applicable.
• Consistent labelling was maintained throughout the dataset.
• Inaccurate or missing annotations were reviewed and corrected.
• Prepared the dataset for POA detection model training.

What made it hard

• Differentiating POA from similar-looking grasses and vegetation.
• Dense and complex field backgrounds made identification harder.
• Partially visible and overlapping plants required careful judgement.
• Keeping bounding boxes accurate for plants of different sizes took attention.
• Consistency had to hold across varied lighting and field conditions.

Result

A high-quality annotated dataset was prepared for POA 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 POA class and was prepared for computer vision model training and agricultural weed detection applications.