Darlan Daison, Ajani Oladayo S, An Joon Woo, Bae Nan Yeon, Lee Bram, Park Tusan, Mallipeddi Rammohan
Department of Artificial Intelligence, Kyungpook National University, Daegu, Republic of Korea.
UBN Corporation, Suseong-gu, Daegu, Republic of Korea.
Sci Rep. 2025 Apr 23;15(1):14019. doi: 10.1038/s41598-025-97168-z.
Agriculture is vital for human sustenance and economic stability, with increasing global food demand necessitating innovative practices. Traditional farming methods have caused significant environmental damage, highlighting the need for sustainable practices like nutrition management. This paper addresses the emerging integration of artificial intelligence (AI) in agriculture, focusing on the specific challenge of growth stage classification of strawberry plants for optimized nutrition management. While AI has been successfully applied in various agricultural domains, such as plant stress detection and growth monitoring, the precise classification of strawberry growth stages remains underexplored. Accurate growth stage identification is vital for timely nutrient application, directly impacting yield and fruit quality. Our research identifies common gaps in existing literature, including limited or inaccessible datasets, outdated methodologies, and insufficient benchmarking. To overcome these shortcomings, we introduce a robust greenhouse-based dataset covering seven distinct strawberry growth stages, captured under diverse conditions. We then benchmark multiple state-of-the-art models on this dataset, finding that EfficientNetB7 achieves a testing accuracy of 0.837-demonstrating the promise of AI-driven approaches for precise and sustainable nutrient management in horticulture.
农业对于人类生存和经济稳定至关重要，随着全球粮食需求的不断增加，创新做法势在必行。传统耕作方法已对环境造成重大破坏，这凸显了营养管理等可持续做法的必要性。本文探讨了人工智能（AI）在农业中的新兴应用，重点关注草莓植株生长阶段分类这一特定挑战，以实现优化营养管理。虽然人工智能已成功应用于植物应激检测和生长监测等多个农业领域，但草莓生长阶段的精确分类仍未得到充分探索。准确识别生长阶段对于及时施肥至关重要，直接影响产量和果实品质。我们的研究发现了现有文献中存在的常见差距，包括数据集有限或难以获取、方法陈旧以及基准测试不足。为克服这些缺点，我们引入了一个强大的基于温室的数据集，涵盖七个不同的草莓生长阶段，这些数据是在不同条件下采集的。然后，我们在这个数据集上对多个先进模型进行基准测试，发现EfficientNetB7的测试准确率达到0.837，这表明人工智能驱动的方法在园艺中实现精确和可持续营养管理方面具有潜力。
