YOLOv8n-ESG: A Framework for Enhanced Strawberry Growth Stage Recognition

Abstract

Abstract Accurate strawberry ripeness detection plays a vital role in quality assurance and market competitiveness enhancement within agricultural production. This study proposes an enhanced YOLOv8n-ESG model for efficient in-field strawberry maturity classification. The methodology categorizes strawberry growth stages into two phases (unripe vs. ripe), addressing quality deterioration risks from improper harvesting timing. Our technical improvements to the baseline YOLOv8n architecture include: 1) backbone network convolution layer optimization, 2) C2f module refinement, 3) attention mechanism integration, 4) loss function modification, and 5) data augmentation implementation. Experimental results demonstrate the model achieves 89.8% precision, 90.5% recall, and 94.8% mAP50 in complex scenarios, effectively mitigating misdiagnosis and missed detection issues. The proposed system enables growers to optimize harvesting schedules through precise ripeness identification, contributing to intelligent agricultural technology development. These advancements in visual recognition systems offer practical solutions for improving postharvest quality control and strengthening market position in perishable fruit supply chains.

Plant phenotyping relevance

イチゴ果実の成熟段階という植物器官の状態を画像認識で推定する手法を開発・評価しており、植物フェノタイピング手法が中心である。

Code and data availability

The paper's strawberry image dataset is stated as not publicly available (available only on request), and no author code, models, or public repository for YOLOv8n-ESG is provided. The only public URL (ultralytics/ultralytics) is the generic baseline framework, not a paper-specific asset.

No evidence-backed public reproduction asset is currently recorded.