Applying machine learning for chili pepper phenotyping and feature extraction

Abstract

Accurate characterization of chili pepper morphology is essential for breeding programs and genetic studies. Traditional phenotyping approaches are often constrained by small sample sizes and a limited set of measurable traits, restricting comprehensive analysis. In this study, we present an automated, image-based phenotyping framework that leverages computer vision and machine learning to extract detailed morphological features from longitudinal slice images of chili peppers. To accurately detect chili fruits and their seeds, the framework employs the YOLOv7 object detection model, achieving a precision of 0.92 and a mean Average Precision (mAP) of 0.87. Building upon these detections, we apply advanced image processing techniques to quantify key phenotypic traits, including seed count, fruit color intensity, length, width, surface area, and surface wrinkle characteristics. These parameters provide critical insights for variety classification, breeding selection, and genetic resource management. The proposed methodology not only enables scalable and reproducible phenotypic assessment but also establishes a searchable dataset of chili pepper varieties, thereby enhancing the efficiency, accuracy, and analytical depth of chili pepper research and breeding programs.

Plant phenotyping relevance

画像と機械学習を用いてトウガラシ果実・種子の形態形質を抽出する枠組みが研究の中心であり、実測精度も評価しているため。

Code and data availability

公開論文であることは確認できましたが、現在の公式API・許可済み取得経路では本文を自動取得できませんでした。

No evidence-backed public reproduction asset is currently recorded.