Sunflower, Ragweed & Grass Segmentation

Project overview

Agricultural field images were annotated for a computer vision dataset focused on segmenting sunflower plants, ragweed and grass in real-world field environments. The project required careful identification and precise polygon segmentation of each target class while distinguishing them from other plants and surrounding vegetation.

The main objective was to create high-quality segmentation data for training an AI model able to recognise and segment different plant types in agricultural fields.

Work performed

• Reviewed and organised the agricultural field image dataset before annotation.
• Sunflower, ragweed and grass plants were identified in each image.
• Precise polygon segmentation masks were created around visible plants.
• Consistent labelling was maintained across all three classes.
• Target plants were carefully separated from surrounding crops, weeds and vegetation.
• Annotated partially visible plants where applicable.
• Every mask followed visible plant boundaries as accurately as possible.
• Quality checks identified and corrected inaccurate or incomplete polygons.
• Confirmed the final annotations were consistent and suitable for segmentation model training.

What made it hard

• Differentiating ragweed and grass from other weeds and field vegetation.
• Sunflower plants surrounded by dense vegetation were hard to isolate.
• Similar appearance between plant species made classification challenging.
• Overlapping leaves and plants made individual segmentation difficult.
• Thin grass structures required careful polygon placement.
• Partially visible plants needed accurate boundary estimation.
• Different growth stages produced significant variation in plant size and appearance.
• Shadows, lighting conditions and complex agricultural backgrounds affected plant visibility.
• Keeping segmentation boundaries accurate across multiple classes required significant attention to detail.

Result

A high-quality sunflower, ragweed and grass segmentation dataset was completed with accurate polygon annotations. The dataset was prepared for training a computer vision model able to identify and segment these plant classes automatically in real-world agricultural environments.

Final deliverable

Accurate and consistent polygon segmentation masks for sunflower, ragweed and grass were delivered throughout the dataset. The final annotations met the project requirements and were ready for AI model training and agricultural crop and weed monitoring applications.