Smart agriculture relies on crop classification, soil analysis, and market information to increase farm efficiency. While conventional techniques like CNN, Random Forest, and SVM exhibit good results, they tend to use big data and lack scalability to meet diverse farming needs. In this paper, we discuss the state-of-the-art of precision agriculture technology and present our proposed approach named Krishi Sarthi, which integrates crop recommendation, soil information, disease identification, and market linkage. Krishi Sarthi is compared with selected state-of-the-art methods, which include early season crop classification (~85% accuracy before June and ~90% by July) through multisensory time series, soil fertility prediction models achieving ~95-97% accuracy, and CNN-based classifiers beating the performances of SVM and Random Forest models with ~95.21% accuracy. Our proposed solution uses React.js and Flask along with a language model called Gemini and CNN/RNN networks to achieve ~94% accuracy in Indian agriculture data sets. In addition, our platform provides timely weather data, market prices, and direct market access for farmers. In conclusion, Krishi Sarthi offers a competitive solution for smart agriculture. Further research will focus on experimental validation, ablation study, and IoT device integration.
Paliwal et al. (Wed,) studied this question.