MobileMold is a comprehensive dataset comprising 4,941 annotated images for food mold detection, captured using smartphones with various clip-on microscope attachments. The dataset addresses the growing need for accessible, low-cost food safety monitoring by leveraging smartphone-based microscopy. This enables research and development in computer vision applications for mold detection on various food surfaces. The mold-detection-baseline repository contains additional resources and scripts for data preprocessing and model training.

Dataset Overview

• Total Images: 4,941
• Annotations: Food Type and Mold Label
• Food Types: 11 categories (carrot, orange, creamcheese, tomato, toast, raspberry, mixed bread, blackberry, blueberry, cheese, onion)
• Microscope Types: 3 different clip-on smartphone microscopes (30x-100x magnification)
• Smartphones: Images captured with 3 different smartphone models

Download

FreshLens Mobile App

The freshlens-app repository contains a Flutter-based mobile app designed for consumer-facing demonstrations and can be used in conjunction with a hosted model. Using a smartphone microscope attachment, users can capture or import images of food. The app then displays the probability that the food is moldy.