Lanata Weed Segmentation

Project overview

Agricultural field images were annotated for a computer vision dataset focused on segmenting Lanata weeds in real-world field environments. Visible Lanata weed plants were identified and precise polygon segmentation masks created to generate high-quality training data for an AI-based segmentation model.

The main objective was to prepare a reliable dataset that could help an AI model accurately identify and segment Lanata weeds while distinguishing them from crops, soil, grass and other surrounding vegetation.

Work performed

• Reviewed and organised the agricultural field image dataset before annotation.
• Visible Lanata weed plants were identified in each image.
• Precise polygon segmentation masks were created along the visible boundaries of the weeds.
• Labelling remained consistent throughout using the Weed class.
• Lanata weeds were carefully distinguished from surrounding crops, grass, soil and other vegetation.
• Partially visible weed plants were annotated where applicable.
• Overlapping leaves and dense vegetation were handled with detailed polygon annotations.
• Quality checks were performed, and inaccurate or incomplete segmentation masks were corrected.
• Annotations were confirmed consistent and suitable for computer vision segmentation model training.

What made it hard

• Differentiating Lanata weeds from other plants and surrounding vegetation.
• Dense field conditions made individual weed boundaries difficult to identify.
• Overlapping leaves and closely positioned plants required careful polygon placement.
• Small or partially visible weeds were challenging to segment accurately.
• Similar-looking vegetation created additional identification challenges.
• Complex backgrounds containing crops, grass, soil and field residue affected visibility.
• Varying lighting conditions and shadows made plant boundaries less clear.
• Precise polygon boundaries had to exclude surrounding vegetation, which demanded significant attention to detail.

Result

A high-quality Lanata weed segmentation dataset was completed with precise polygon annotations under the Weed class. The dataset was prepared for training a computer vision model capable of automatically identifying and segmenting Lanata weeds in agricultural field environments.

Final deliverable

Accurate and consistent polygon segmentation masks for Lanata weeds were delivered throughout the dataset. The final annotations were prepared according to project requirements and were ready for AI model training and automated weed detection and monitoring applications.