Development of maize plant dataset for intelligent recognition and weed control

Area of Science:

• Agricultural Science
• Computer Vision
• Machine Learning

Background:

• Accurate identification of maize plants and weed species is essential for effective weed management in agriculture.
• Automated herbicide application requires high-quality datasets for training recognition models.

Purpose of the Study:

• To develop a comprehensive dataset for maize plant and weed recognition.
• To support the precise automated application of herbicides.
• To establish a benchmark for computer vision and machine learning in intelligent agriculture.

Main Methods:

• Collected 36,374 high-resolution images of maize plants and associated weeds from 18 locations in North Central Nigeria.
• Annotated 500 images using the Labelimg suite for detailed feature extraction.
• Focused on capturing diverse conditions representative of farmland environments.

Main Results:

• A large-scale, annotated dataset specifically for maize and weed identification has been created.
• The dataset encompasses a variety of weed species commonly found in maize fields.
• The data facilitates the development and testing of advanced recognition algorithms.

Conclusions:

• The developed maize plant and weed dataset is a valuable resource for agricultural AI research.
• It enables advancements in automated weed detection and targeted herbicide spraying.
• This benchmark dataset will accelerate innovation in intelligent farming systems.