A large-scale, data-centric review has mapped the landscape of computer vision datasets that underpin artificial intelligence in modern agriculture. The study, published in Artificial Intelligence Review, used a PRISMA-guided systematic review methodology to examine how agricultural datasets are built, annotated and shared, and to identify the structural weaknesses that limit their usefulness for developing reliable and generalizable agricultural AI systems.
The researchers conducted a systematic search of the peer-reviewed data journals Scientific Data and Data in Brief, with the search completed through 31 January 2026. The search identified 2,308 records, from which 253 eligible dataset publications were retained after duplicate removal and eligibility screening based on predefined inclusion and exclusion criteria. This filtering process ensured that the analysis focused on dataset publications meeting consistent quality and relevance standards.
The review analyzed the dataset landscape across multiple dimensions, including vision tasks, annotation strategies, acquisition environments, sensing modalities, and geographic distribution. By examining these dimensions together, the authors were able to characterize how agricultural computer vision datasets are currently designed and where their coverage falls short.
A central contribution of the study is a unified multi-dimensional taxonomy that organizes datasets based on task orientation, semantic content, and design characteristics. The taxonomy is intended to enable structured comparison and informed dataset selection, giving researchers and practitioners a systematic way to identify which datasets suit particular agricultural vision tasks.
The analysis identified key limitations that collectively hinder reproducibility and model generalization. These include sampling bias, underrepresentation of crops and regions, inconsistent annotation practices, and limited interoperability. Such weaknesses mean that models trained on available datasets may not perform reliably when applied to new crops, regions or farming conditions.
Beyond these dataset-level limitations, the review also highlighted challenges in dataset accessibility and the absence of standardized benchmarking frameworks, which restrict effective knowledge transfer across the research community. Without common benchmarks, it is difficult to compare methods fairly or to build cumulatively on previous work.
To address these gaps, the authors outline critical directions for future research, emphasizing standardized dataset design, scalable annotation frameworks, and robust evaluation protocols. These directions are framed as necessary steps to support the development of reliable and generalizable agricultural AI systems.
The study’s data-centric approach reflects the growing importance of dataset quality in AI-driven agriculture, where automated analysis increasingly depends on the availability of well-constructed computer vision resources. By documenting trends, proposing a taxonomy and specifying future directions, the review provides a structured foundation for improving how agricultural datasets are created and evaluated.
The article is published open access, and the version shared carries a permanent DOI as peer-reviewed, accepted research that is citable while remaining subject to further edits before the final Version of Record.
Subject of Research: A systematic review of agricultural computer vision datasets, their taxonomy, limitations and future research directions
Article Title: The landscape of computer vision datasets in smart agriculture: trends, taxonomies, challenges, and future directions
Article References: Morshed, M. N., Maherin, S., Islam, M. M., & Islam, M. S. (2026). The landscape of computer vision datasets in smart agriculture: trends, taxonomies, challenges, and future directions. Artificial Intelligence Review. https://doi.org/10.1007/s10462-026-11754-9
Image Credits: AI Generated
DOI: 10.1007/s10462-026-11754-9
Keywords: smart agriculture, computer vision, datasets, systematic review, PRISMA, annotation, taxonomy, benchmarking, reproducibility, agricultural AI, landscape, computer
News Source: Denise Maddox. (November 5, 2026). Agricultural AI datasets mapped in sweeping review of 253 studies. Scienmag.
