LeaData a novel reference data of leather images for automatic species identification

Area of Science:

• Materials Science
• Computer Vision
• Biotechnology

Background:

• Traditional leather identification methods lack accuracy and objectivity.
• Accurate species identification is crucial for the leather industry, consumer protection, and biodiversity preservation.
• Leveraging big data and image analysis offers a path toward automated and precise leather identification.

Purpose of the Study:

• To create a novel and comprehensive leather image dataset (LeaData) for automated species identification.
• To enable objective analysis of leather grain surfaces for accurate species determination.
• To facilitate the development of accessible smart identification techniques for the leather industry.

Main Methods:

• Acquisition of leather images using a simple, handheld digital microscope at 47× magnification.
• Collection of 38,172 images across four mammalian species (buffalo, cow, goat, sheep) from 137 leather samples.
• Inclusion of diverse grain patterns, including ideal, non-ideal, and variations across body parts.

Main Results:

• Development of LeaData, a unique and extensive collection of leather images.
• Demonstration of the potential for automated species identification through grain surface pattern analysis.
• Establishment of a foundational dataset for advancing digital image processing in leather technology.

Conclusions:

• LeaData provides a robust resource for developing smart, objective leather species identification systems.
• This digitized approach supports digitization in leather technology, enhancing traceability and quality control.
• The dataset aids in biodiversity preservation and consumer protection by enabling reliable species authentication.