Indian agriculture is confronted by persistent challenges that undermine smallholder productivity: late identification of crop diseases, empirical and often wasteful fertilizer practices, and informational asymmetry in commodity markets. This paper presents CropDoc AI, an integrated intelligent decision support system designed to address these interlinked challenges within a single, accessible platform. The system employs a convolutional neural network (CNN) trained on the PlantVillage benchmark dataset to classify crop leaf photographs into disease and nutrient deficiency categories, achieving a top-1 classification accuracy of approximately 95%. A FastAPI-based inference server exposes the trained model as a RESTful endpoint, receiving uploaded leaf images and returning structured diagnostic predictions in near real time. Diagnostic outputs are coupled with an adaptive fertilizer recommendation module that fuses identified nutrient deficiencies with current and seven-day weather forecasts retrieved via the OpenWeatherMap API, generating context-sensitive agrochemical guidance. Simultaneously, a market intelligence module retrieves live commodity prices from the e-NAM and Agmarknet portals, enabling farmers to make informed harvest-timing and selling decisions. Financial calculators embedded within the platform support yield estimation, investment planning, break-even analysis, and return-on-investment projection. The entire system is packaged as a cross-platform desktop application using the Electron framework, ensuring offline-tolerant local operation without dependence on cloud subscriptions. Evaluation demonstrates that the integrated system meaningfully reduces fertilizer over-application, improves disease response speed, and provides price discovery that was previously inaccessible to rural users. CropDoc AI represents a practical step toward evidence-driven, technology-assisted smallholder farming in the Indian context.
K.Thanmay et al. (Fri,) studied this question.