Upper and Lower Leaf Side Detection with Machine Learning Methods.

Area of Science:

• Botany
• Plant Pathology
• Computer Science

Background:

• Current AI disease identification often uses only one leaf side.
• This limits accuracy as diseases can present differently on upper and lower leaf surfaces.
• Distinguishing leaf sides is crucial for comprehensive phytopathological analysis.

Purpose of the Study:

• To develop a classification model for identifying the upper and lower sides of plant leaves.
• To leverage distinct visual features of leaf sides for improved AI-driven disease detection.
• To enhance the accuracy of foliar plant disease identification systems.

Main Methods:

• Utilized botanical knowledge to identify key differentiating features between leaf sides.
• Investigated color variations due to sun exposure as a classification indicator.
• Analyzed the prominence of leaf veins and the curvature (concave/convex) of leaf surfaces.
• Applied both deep learning and traditional machine learning models for classification.

Main Results:

• Identified color, vein prominence, and leaf curvature as reliable features for distinguishing leaf sides.
• Demonstrated the feasibility of classifying leaf sides using these botanical characteristics.
• Laid the groundwork for integrating dual-sided leaf analysis into plant disease identification.

Conclusions:

• Classifying leaf sides is a vital step towards more accurate AI-based plant disease diagnosis.
• The proposed method using botanical features offers a robust approach to leaf side identification.
• Future work can integrate this classification into advanced phytopathological AI tools.