The application of deep learning techniques in agriculture has gained significant attention due to their potential to improve crop monitoring, classification, and yield forecasting. However, evidence regarding their adoption, performance, and operational suitability within Sub-Saharan Africa remains fragmented. This study conducted a systematic literature review to synthesize existing evidence on deep learning architectures for crop type mapping and yield prediction in Sub-Saharan Africa. The review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020) framework. Literature was retrieved from Scopus, Google Scholar, Frontiers, and MDPI databases using predefined search strings. An initial 4,777 records were identified, from which 27 studies met the inclusion criteria and were subjected to qualitative synthesis. The review examined deep learning architectures employed, remote sensing and geospatial datasets utilized, model performance, and emerging trends in agricultural monitoring. The findings revealed that Convolutional Neural Networks (CNNs), Artificial Neural Networks (ANNs), Long Short-Term Memory (LSTM) networks, Deep Neural Networks (DNNs), transfer learning frameworks, and hybrid models were the dominant architectures used for crop type mapping and yield prediction. Sentinel-2, Landsat, MODIS, UAV imagery, climatic datasets, and soil information constituted the primary data sources. The review further showed that deep learning models achieved high classification accuracies and strong predictive performance, particularly when multiple datasets were integrated through data fusion techniques. Emerging trends identified include transfer learning, explainable artificial intelligence, hybrid deep learning architectures, attention-based models, UAV-assisted monitoring, and IoT-enabled agricultural systems. The study concludes that deep learning technologies offer substantial potential for improving agricultural monitoring and yield forecasting across Sub-Saharan Africa. However, challenges relating to data scarcity, model transferability, and unequal geographical coverage remain. The study recommends increased investment in open agricultural datasets, multi-source data integration, and explainable artificial intelligence frameworks to enhance the scalability and operational deployment of deep learning systems in the region.
Bolaji et al. (Wed,) studied this question.