Grocery Item Counter using Computer Vision

This repository contains a working prototype for an application that counts items in a grocery shop or inventory warehouse using a camera. The system leverages deep learning (using a pre-trained YOLOv5 model) to detect objects, counts them, and associates each detection with product details from a simulated product database.

Table of Contents

• Overview
• Features
• Project Structure
• Prerequisites
• Installation
• Running the Prototype
• Deployment
  • Containerization with Docker
  • Cloud Deployment
  • Production Considerations
• Data Engineering & Deep Learning
• Contributing
• License

Overview

This project demonstrates an end-to-end pipeline for counting items using a camera. The application:

• Captures images via a webcam.
• Processes images with a YOLOv5 model (loaded via PyTorch Hub) to detect and count objects.
• Looks up associated product details (brand and product name) using a simulated product database.
• Displays the detection results with bounding boxes and labels.

The system is designed to function even if detailed product information is not available—by simply aggregating the count of similar items.

Features

• Real-time Image Capture: Uses OpenCV to capture images from a webcam.
• Object Detection: Utilizes YOLOv5 to identify items in the captured image.
• Product Counting: Counts detected items and aggregates results.
• Product Lookup: Matches detected labels with a product details dictionary.
• Visualization: Annotates and displays images with bounding boxes and labels.

Project Structure

• main.py: Main script to capture an image, perform object detection, count items, and display annotated results.
• README.md: This file.
• requirements.txt: (Optional) File listing project dependencies.

Prerequisites

• Python 3.7 or later
• A functional webcam connected to your system

Python Dependencies

The project relies on the following libraries:

• OpenCV (opencv-python)
• PyTorch (torch and torchvision)
• NumPy (numpy)