SERICARE AI

Sericulture, the practice of silk production through silkworm rearing, remains highly vulnerable to viral, fungal, and microsporidian diseases such as Grasserie, Pebrine, Flacherie, and Muscardine, which cause substantial economic losses to farmers due to delayed or incorrect diagnosis.Existing approaches to automated silkworm disease identification largely stop at image classification and do not connect a diagnosis to actionable, farmer-usable guidance.This paper presents SeriCare AI, an integrated web-based system that combines a custom Convolutional Neural Network (CNN) for silkworm disease classification with a Retrieval-Augmented Generation (RAG) based recommendation module and an automatically generated Digital Health Card.The system is built on a React frontend and a Flask/MongoDB backend, and is designed for low-friction use by farmers rather than as a data-science dashboard.The current implementation targets binary classification between Healthy and Grasserie-infected silkworms using a dataset of approximately 8,700 labelled images, with planned expansion to Pebrine, Flacherie, and Muscardine for full five-class coverage.Unlike prior work that treats detection as a standalone task, SeriCare AI's primary contribution is the end-to-end pipeline that couples classification output with disease-specific treatment and environmental recommendations and persists this information as a health record for the farmer.This paper describes the system architecture, dataset, CNN design, and evaluation protocol, and reports the status of ongoing experimental validation.