AgroGuard AI: A Zero-Shot Multimodal Crop Disease Diagnostic Assistant for Smallholder Farming

Abstract

Smallholder farmers in India frequently incur avoidable agricultural input costs due to non-diagnostic decision- making — replicating a neighboring farmer’s treat- ment schedule or acting on retailer-driven upselling — rather than an actual, verified assessment of their crop’s health, a behavioral pattern that often pushes vulnera- ble farmers toward avoidable debt. Existing plant dis- ease detection systems, while accurate in controlled set- tings, remain research prototypes that assume technical literacy and require dataset-specific training, offering no low-friction mechanism to interrupt this cycle. This work presents AgroGuard AI, a lightweight, browser- based diagnostic assistant that enables a farmer to inde- pendently verify crop health using only a photograph, requiring no typing, technical training, or intermedi- ary. The system encodes the captured image and a role-conditioning prompt into a single multimodal re- quest processed by Google’s Gemini API, which per- forms zero-shot visual reasoning — via patch-based image segmentation and self-attention — to generate a plain-language diagnosis, confidence estimate, and treatment recommendation. Evaluation on a small set of test images demonstrated preliminary diagnos- tic capability without any dataset-specific fine-tuning. Critically, this work identifies and explicitly character- izes an important epistemic distinction largely over- looked in similar systems: the model’s stated con- fidence percentage is a generated linguistic estimate, not a calibrated statistical probability as in a closed- set classifier. While limitations around internet de- pendency, sample size, and visual explainability re- main, the system demonstrates that accessible, mean- ingful agricultural decision support can be achieved without heavy infrastructure or custom training data.