FoodSense AI Sensor

Version 1.0.0

Objective

This research project focuses on the development of a smart chemical sensor sticker for visual detection of food spoilage integrated with artificial intelligence. The system uses a chemical colorimetric sensor that undergoes visible color changes when exposed to volatile organic compounds (VOCs) and biogenic amines produced during the food decomposition process.

Methodology

The system employs a Convolutional Neural Network (CNN) based on the MobileNetV2 architecture with transfer learning from ImageNet. The model is trained on a dataset of chemical sensor images captured at various stages of food spoilage, enabling accurate classification into four distinct levels:

• Fresh — No detectable signs of spoilage
• Safe — Minor early changes; food is still safe
• Warning — Early spoilage indicators detected
• Spoiled — Significant decomposition; not safe for consumption

System Architecture

The platform is built using a professional Flask Blueprint architecture with a PostgreSQL database backend, containerized with Docker for reproducible deployment. The AI inference pipeline processes images in real-time, providing instant classification results with confidence scores and probability distributions.

Applications

This technology has potential applications in food safety monitoring across the supply chain, from production facilities to retail and consumer households. The non-invasive, low-cost sensor combined with smartphone-based AI analysis makes it accessible for widespread deployment in food quality assurance.