ShelfSense AI

Inspiration

Retail stores lose significant revenue due to empty shelves, misplaced products, inaccurate inventory counts, and delayed replenishment. Employees spend valuable time manually checking shelves instead of assisting customers.

We wanted to build an AI system that continuously monitors store shelves, detects inventory issues instantly, and provides actionable recommendations before sales are lost.

What it does

ShelfSense AI transforms ordinary security cameras into intelligent retail assistants.

Using computer vision and AI agents, the platform continuously analyzes shelf images to identify out-of-stock products, misplaced items, pricing inconsistencies, and low inventory levels.

Managers receive real-time alerts, demand forecasts, and automated restocking recommendations through an intuitive dashboard.

The platform also provides sales analytics, inventory insights, and AI-generated recommendations to improve store operations.

Key Features

Real-time shelf monitoring
Computer vision inventory detection
Automatic out-of-stock alerts
Product recognition
Smart replenishment recommendations
Demand forecasting
AI inventory assistant
Retail analytics dashboard

How we built it

The frontend was built using Next.js, React, and Tailwind CSS to provide an intuitive dashboard for retail managers.

The backend uses FastAPI with PostgreSQL and Redis for data processing and caching.

Computer vision models analyze shelf images to detect products and estimate inventory levels. AI agents coordinate image processing, inventory validation, forecasting, and recommendation generation.

OpenAI powers the conversational retail assistant, allowing managers to ask questions like "Which products need restocking today?" or "What items are selling faster than expected?"

Challenges we ran into

Lighting conditions, camera angles, and partially hidden products made object detection difficult. We improved reliability by combining object detection with historical inventory data and confidence scoring.

Another challenge was generating recommendations that balanced inventory availability with predicted customer demand.

Accomplishments that we're proud of

Built a real-time AI shelf monitoring system.
Automated inventory checking without additional hardware.
Developed AI-powered demand forecasting.
Created a conversational retail assistant for store managers.

What we learned

Building computer vision systems for retail requires more than object detection. Combining visual recognition with AI reasoning and historical sales data produces much more useful recommendations.