Using comprehensive machine-learning models to classify complex morphological characters.

Area of Science:

• Morphological characterization
• Machine learning applications in biology
• Bioinformatics

Background:

• Quantitatively measuring complex morphological characters is challenging.
• Understanding morphological variability is key to studying phenotypic evolution.
• Existing methods may require large datasets or significant computational resources.

Purpose of the Study:

• To develop a machine learning pipeline (SVMorph) for classifying complex morphological characters.
• To provide a method applicable to both small and large biological datasets.
• To offer an efficient and reliable tool for non-model organisms.

Main Methods:

• Integrated Histogram of Oriented Gradient (HOG) and Local Binary Pattern (LBP) descriptors.
• Implemented image data augmentation to enhance feature extraction and model generalizability.
• Utilized a Support Vector Machine (SVM) classifier within the pipeline.

Main Results:

• SVMorph demonstrated reliability and speed in texture-based individual classification.
• The pipeline performed effectively on small training datasets and with limited computational power.
• Comparative analysis showed SVMorph's advantages over traditional techniques and some Convolutional Neural Network (CNN)-based methods.

Conclusions:

• SVMorph offers an efficient solution for classifying multiple morphological characters in diverse organisms.
• The pipeline is suitable for studies with limited data or computational constraints.
• SVMorph facilitates the study of phenotypic evolution and variability in non-model organisms.