Machine learning & AI

FreshLens

AI-powered inventory and freshness monitoring system for small-scale retailers.

Abstract

FreshLens is an AI-powered inventory and freshness monitoring system designed for small-scale retailers. It enables vendors to capture produce images through a mobile application, automatically classify produce freshness, monitor inventory, and receive alerts for low-stock, aging, and spoilage conditions. A web-based platform allows administrators to manage vendors, products, shelf-life configurations, alerts, and platform analytics.

The problem

Small-scale retailers often need to monitor the quantity and freshness of perishable produce manually. This can make it difficult to identify aging or spoiled items early, maintain accurate inventory information, and respond quickly to low-stock conditions.

FreshLens was proposed to provide an automated system that combines inventory monitoring with AI-based freshness classification, helping vendors make better decisions about their produce inventory.

The solution

FreshLens provides a vendor mobile application where vendors can capture produce images, confirm quantities, submit scans, and view freshness and inventory information.

The system processes submitted images asynchronously using a two-tier CNN-based freshness classification pipeline. The system identifies produce and determines its freshness state, while inventory and batch information are used to support low-stock and aging alerts.

Vendors receive push notifications when scan results or relevant alerts become available.

A separate platform administrator web application provides functionality for managing vendors, product catalogues, shelf-life configurations, alerts, and platform analytics.

In detail

FreshLens is an AI-powered inventory and freshness monitoring system designed for small-scale retailers. The platform consists of a React Native/Expo vendor mobile application and a Next.js administrator web application, communicating with a FastAPI REST backend.

Vendors can capture produce images, confirm quantities, submit scans, and monitor freshness and inventory information. Submitted scans are processed asynchronously using Redis and Celery through a two-tier CNN pipeline for produce identification and freshness classification. Results are stored in PostgreSQL, while produce images are stored in Cloudflare R2.

The system generates low-stock, aging, and spoilage alerts and delivers relevant notifications to vendors through mobile push notifications. PostgreSQL Row-Level Security (RLS) is used to enforce tenant isolation and protect vendor-specific data.

The platform also provides administrators with vendor management, product catalogue management, shelf-life configuration, alert management, and platform analytics.

Related projects

Machine learning & AI

DebtLens

AI-powered SaaS platform for detecting, predicting and prioritizing technical debt in GitHub repositories.

188
Machine learning & AI

SmartJourney

AI-powered multi-agent tourism assistant for personalized trip planning and travel recommendations.

87 2026