Artificial Intelligence Adoption in Smart Agriculture: A Review of Convolutional Neural Networks for Plant Disease Detection and Agribusiness Sustainability
Anggi Oktaviani *
Department of Informatics, Universitas Nusa Mandiri, Jakarta, Indonesia.
Dahlia Sarkawi
Department of Office Administration, Universitas Bina Sarana Informatika, Jakarta, Indonesia.
Agus Priadi
Department of English Literature, Universitas Bina Sarana Informatika, Jakarta, Indonesia.
*Author to whom correspondence should be addressed.
Abstract
Plant diseases remain one of the most persistent and economically damaging threats to global food security, with yield losses across major staple crops running into the tens of billions of dollars each year. The rise of artificial intelligence, and convolutional neural networks (CNNs) in particular, has opened genuinely new possibilities for detecting plant disease early, accurately, and at scale. This review critically examines the state of CNN-based plant disease detection within the wider context of smart agriculture and agribusiness sustainability. Drawing on peer-reviewed literature published between January 2016 to February 2026, the paper traces the evolution of CNN architectures, training methods, and benchmark performance across a wide range of crops and disease categories. Particular attention is given to transfer learning, data augmentation, and lightweight architecture design as responses to the recurring problem of limited annotated training data. The paper also considers how CNNs are being combined with complementary technologies, including the Internet of Things, unmanned aerial vehicles, and edge computing, and what this means for deployment in real farming conditions. Economic and sustainability dimensions are explored throughout, with attention to whether the gains from AI adoption are likely to reach smallholder farmers or remain concentrated among larger, better-resourced agribusinesses. Despite genuinely impressive results under controlled benchmark conditions, several barriers to field deployment persist: dataset bias, poor generalisation in complex agricultural environments, computational constraints, and a continuing shortfall in model interpretability. The review closes by identifying priority research directions, including cross-domain transfer learning, explainable AI, the development of field-representative datasets, and participatory approaches to tool design. Taken together, the evidence suggests that CNN-based disease detection holds real promise for agribusiness sustainability, but realising that promise will depend on sustained interdisciplinary collaboration and deployment strategies that are sensitive to local context rather than assuming one-size-fits-all solutions.
Keywords: Convolutional neural networks, plant disease detection, smart agriculture, precision agriculture, deep learning, transfer learning, agribusiness sustainability, internet of things, unmanned aerial vehicles, food security