A Deep Driven Smart Farming Systems for Sustainable Agriculture

Abstract

Agricultural productivity is very sensitive to temporal variations in climate, heterogeneous soil conditions, and dynamic crop health. As such, precise crop yield forecasting constitutes a complex and multifactorial problem. Long-term temporal dependencies in agricultural datasets cannot be modeled by conventional statistical and machine learning approaches. An advanced deep learning-based framework for intelligent agricultural analytics presented this study integrates crop yield prediction with crop health assessment using multi-source datasets. Time series data that include meteorological parameters, soil characteristics, and historical yield records are modeled with Temporal Convolutional Networks (TCNs) to learn long-range seasonal dependencies effectively as well as temporal trends. For visual analysis of crops, both satellite imagery and leaf-level photographs go through a Vision Transformer (ViT) architecture wherein self-attention mechanisms are used to extract global spatial features so that subtle patterns related to the health of crops can be identified along with diseases affecting them. The framework uses publicly available datasets: the CY-Bench crop yield benchmark for predictive modeling; Indian historical crop yield and weather datasets for temporal analysis; and the PlantVillage image dataset for disease detection and visual feature extraction. Standard evaluation metrics such as accuracy, mean absolute error (MAE), and root mean square error (RMSE) will be used to assess performance. Results from experiments show that the proposed deep learning framework with TCNs plus Vision Transformers outperforms traditional models in capturing complex spatiotemporal patterns—hence improving accuracy plus reliability in both crop yield prediction as well as monitoring crop health.

Introduction

The study proposes an intelligent smart farming framework that integrates temporal and image-based data to enhance crop yield prediction and crop health monitoring. Traditional statistical and machine learning models struggle with long-term temporal dependencies and high-dimensional spatial relationships in agricultural data. To address this, the framework combines Temporal Convolutional Networks (TCNs) for modeling sequential time-series data (weather, soil, historical yields) and Vision Transformers (ViTs) for extracting global spatial features from plant leaf and satellite images.

Key components include:

1. Data preprocessing and alignment to handle heterogeneous datasets (CY-Bench, Indian agricultural datasets, PlantVillage).
2. Feature extraction and modeling, where TCN captures seasonal trends and ViT identifies crop health patterns and diseases.
3. Integration of model outputs for combined crop yield prediction and health classification.
4. Evaluation using metrics like Accuracy, RMSE, MAE, and F1-score to validate performance.

The framework offers a unified deep learning solution capable of capturing both spatiotemporal dependencies, making it suitable for smart, sustainable, and scalable agriculture.

Conclusion

This paper has presented an intelligent smart farming system using TCNs and ViTs to ensure comprehensive analysis of the data. The intelligent system has effectively integrated time-series data, including weather, soil, and past crop yields, with images of plant leaves. This ensures effective crop yield predictions. The experiment has shown that the TCN model has effectively captured long-term temporal dependencies to ensure accurate predictions of crop yields under different environmental conditions. In addition, the Vision Transformer has effectively captured global spatial features from images to ensure high accuracy in disease detection from images. This has been achieved with an accuracy of 92.8%, with high precision, recall, and F1-score values. The use of multi-modal data will enhance the ability of the system to make generalizations over different agricultural scenarios. The low values of MAE and RMSE validate the robustness and consistency of the suggested approach for crop yield prediction. In conclusion, the suggested hybrid deep learning model is better than the conventional machine learning approach and the conventional deep learning approach due to the ability to incorporate both temporal and spatial trends. The system will serve as an efficient tool for precision agriculture.

References

[1] D. Tripathi and S. K. Biswas, “Design of a precise ensemble expert system for crop yield prediction using machine learning analytics,” Journal of Forecasting, vol. 43, no. 8, pp. 3161–3176, 2024. [2] Elavarasan and P. M. D. Vincent, “Crop yield prediction using deep reinforcement learning model for sustainable agrarian applications,” IEEE Access, vol. 8, pp. 86 886–86 901, 2020. [3] Bai, S., Kolter, J. Z., & Koltun, V., “An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling,” arXiv preprint arXiv:1803.01271, 2018. [4] Singh, P. T., et al., “Explainable Vision Transformer Enabled CNN for Plant Disease Identification (PlantXViT),” arXiv, 2022. [5] Tonmoy, M. R., et al., “MobilePlantViT: A Mobile-Friendly Hybrid Vision Transformer for Plant Disease Classification,” arXiv, 2025. [6] Upadhyay, A., et al., “Deep Learning and Computer Vision in Plant Disease Detection: A Comprehensive Review,” Artificial Intelligence Review (Springer), 2024. [7] Thakur, P. S., et al., “A Deep Learning-Based Approach for Automated Plant Disease Classification Using Vision Transformer,” Scientific Reports (Nature), 2022. [8] Rahman, M., et al., “AI-Driven Smart Agriculture Using Hybrid Transformer-CNN Models,” Scientific Reports (Nature), 2025. [9] Ahmad, A., et al., “TrIncNet: A Lightweight Vision Transformer Network for Plant Disease Identification,” Frontiers in Plant Science, 2023. [10] Mohameth, F., et al., “Plant Disease Detection with Deep Learning and Feature Extraction Using PlantVillage Dataset,” Journal of Computer and Communications, 2020. [11] Recent Advances in CNN-Based Plant Disease Detection, International Journal of Recent Development in Engineering and Technology, 2026. [12] Roy, A. M., et al., “A Fast and Accurate Object Detection Model for Plant Disease Detection Using YOLOv4,” arXiv, 2021. [13] Mohanty, S. P., Hughes, D. P., & Salathé, M., “Using Deep Learning for Image-Based Plant Disease Detection,” Frontiers in Plant Science, vol. 7, 2016. [14] Dosovitskiy, A., et al., “An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale,” International Conference on Learning Representations (ICLR), 2021. [15] Modi A., Sharma P., Saraswat D., Mehta R. Review of crop yield estimation using machine learning and deep learning techniques. Scalable Computing. 2022;23(2). [16] van Klompenburg T., Kassahun A., Catal C. Crop yield prediction using machine learning: a systematic literature review. Comput. Electron. Agric. 2020;177(July). [17] Filippi P. 0123456789; 2019. An Approach to Forecast Grain Crop Yield Using Multi - Layered , Multi - Farm Data Sets and Machine Learning. [18] U. Bhimavarapu, G. Battineni, N. ChintalapudiImproved Optimization Algorithm in LSTM to Predict Crop Yield 2023. [19] Azuaje F. Review of \" data mining : practical machine learning tools and techniques \" by witten and frank. January 2006. 2014. [20] El-Kenawy, E. S. M., Alhussan, A. A., Khodadadi, N., Mirjalili, S., & Eid, M. M. (2024). Predicting potato crop yield with machine learning and deep learning for sustainable agriculture. Potato Research. [21] M. Ashfaq, I. Khan, A. Alzahrani, M.U. Tariq, H. Khan, A. GhaniAccurate Wheat Yield Prediction Using Machine Learning and Climate-NDVI Data Fusion IEEe Access., 12 (2024), pp. 40947-40961, 10.1109/ACCESS.2024.3376735 [22] L.S. Cedric, W.Y.H. Adoni, R. Aworka, J.T. Zoueu, F.K. Mutombo, M. Krichen, C.L.M. KimpoloCrops yield prediction based on machine learning models: case of West African countriesSmart Agricultural Technology, 2 (2022), Article 100049, 10.1016/j.atech.2022.100049

Copyright

Copyright © 2026 V. Chaitanya Harsha Vardhan, S. Praveen, Sk. Abdul Sattar, Sk. Jani Babu, Yerininti Venkata Narayana. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.