Comprehensive data of 10 fruit leaf classes captured for agricultural AI applications

Area of Science:

• Agricultural Science
• Computer Vision
• Data Science

Background:

• South Asia faces a scarcity of high-quality, region-specific agricultural image datasets.
• Existing datasets may lack the resolution and standardization needed for advanced machine learning applications.

Purpose of the Study:

• To create a comprehensive, high-resolution dataset of fruit tree leaves from Bangladesh.
• To support the development of computer vision models for agricultural applications in the region.

Main Methods:

• Collected 3173 images of leaves from ten fruit species using Realme 7-Pro and 8-Pro smartphones.
• Photographed leaves under controlled lighting against a white background, capturing both top and underside views.
• Ensured a balanced class distribution (300-343 images per species) for machine learning suitability.

Main Results:

• A publicly available dataset with standardized, high-resolution images of fruit leaves.
• The dataset includes diverse visual properties like leaf shape, venation, and surface texture.
• Balanced class distribution suitable for training machine learning models.

Conclusions:

• The dataset addresses the need for South Asian agricultural image data.
• Facilitates the development of automated crop recognition and monitoring systems.
• Enables advancements in agricultural computer vision research and applications.