Chilean Needle Grass Segmentation

Project overview

Agricultural field images were annotated for a computer vision dataset focused on detecting and segmenting Chilean needle grass in natural and agricultural environments. The work involved creating precise polygon annotations around visible grass plants to generate high-quality training data for an AI-based segmentation model.

Work performed

• Reviewed and organised the field image dataset before annotation began.
• Identified visible Chilean needle grass plants in each image.
• Created precise polygon annotations following the visible boundaries of the target grass.
• Maintained consistent labelling using the Chilean_Needle_Grass class.
• Partially visible grass plants were annotated wherever applicable.
• The target grass was distinguished carefully from surrounding crops, weeds, dry grass and other vegetation.
• Thin leaves and irregular plant structures were handled with detailed polygon annotations.
• Quality checks were carried out and inaccurate or incomplete polygons were corrected.
• Annotations were confirmed as consistent and suitable for training a segmentation model.

What made it hard

• Chilean needle grass had thin, narrow leaves that were difficult to outline precisely.
• Dense vegetation made individual plants difficult to separate.
• Similar-looking grasses and weeds complicated identification.
• Overlapping plants required careful polygon placement.
• Partially visible plants needed accurate boundary estimation.
• Different lighting conditions, shadows and field backgrounds affected visibility.
• Precise polygon boundaries around thin plant structures required significant attention to detail.

Result

A high-quality Chilean needle grass segmentation dataset was completed with precise polygon annotations. The dataset was prepared for training a computer vision model capable of detecting and segmenting Chilean needle grass automatically in real-world field environments.

Final deliverable

Accurate and consistent polygon annotations for Chilean needle grass were delivered throughout the dataset. The final output met the project requirements and was ready for AI model training and agricultural vegetation monitoring applications.