Abstract : This paper presents a deep learning-based solution developed to detect silkworm diseases and optimize rearing environments using image inputs. The system, called Silk Shield, utilizes a fine-tuned EfficientNetB3 model capable of classifying silkworm conditions including Pebrine, Grasserie, Muscardine, and Flacherie by analyzing physical traits from images. It also estimates optimal temperature and humidity for cocoon development using climate-aware intelligence. Unlike conventional systems requiring manual inspection or external sensors, Silk Shield uses vision-based learning, making it highly scalable and affordable. With high prediction accuracy across categories, it is integrated into a real-time interface that provides farmers with quick, actionable insights. This work highlights how AI can transform sericulture into a smarter, sensor-free process.

| DOI: 10.17148/IARJSET.2025.125345