FURI Project Progress

Project Update: Automating Yard Management with Object Detection and OCR

Project Description

This project aims to automate the Yard Management System (YMS) by using Machine Learning (ML) technologies, particularly Object Detection (OD) and Optical Character Recognition (OCR). In the traditional process, when trucks enter or exit a yard, a gate operator manually checks the container details and records them into a system, which can take anywhere from 5 to 10 minutes per truck.

This project leverages camera-based object detection and OCR to automatically capture container information, streamlining the process and reducing the time per truck to as little as 10 seconds. This automation eliminates manual errors, increases efficiency, and provides a modern solution to yard management.

Technology Stack

  • Frontend: React, JavaScript, TypeScript
  • Framework: Next.js
  • Backend: Node.js, Python, Supabase (BaaS Platform)
  • Deployment: Vercel, AWS EC2 with Load Balancer, Route53

Project Repository

Weekly Progress

Week 1-4: Initial Setup and Research

  • Project Initiation: Defined project goals, team roles, and responsibilities.
  • Research: Investigated challenges with traditional yard management and explored Object Detection (OD) and OCR technologies.
  • System Design: Planned the overall architecture of the system and drafted UI/UX design sketches for the frontend interface.

Week 5-8: Prototype and Initial Development

  • Prototype Development: Integrated YOLOv5 (a deep learning model for object detection) to detect container numbers in sample images.
  • OCR Integration: Implemented AWS Textract to convert images into text (container numbers and sizes).
  • Frontend Development: Developed the main page, including functionality to display detected container images and added video download options for users to view the object detection process.

Week 9-12: Integration and Testing

  • Backend Integration: Implemented real-time data storage using Node.js and Supabase.
  • Data Handling: Created a process to store container data (e.g., container number, size, date, and driver’s name) in the Supabase database.
  • Testing: Improved the object detection model, integrated image storage on AWS, and set up features to display detected images in the frontend.

Week 13 and Beyond: Full System Deployment

  • Real-Time Video and Object Detection: Enabled users to download videos and observe how the system detects containers.
  • UI Enhancements: Refined the main page to display detected images and ensure a smooth user experience.
  • Deployment: Deployed the project on Vercel and AWS, ensuring scalability and reliability for production use.

How it Works

  1. Camera Input: A camera installed at the gate captures images of incoming or outgoing containers.
  2. Object Detection: YOLOv5 is used to identify and extract the container number from the images.
  3. OCR Processing: The detected images are sent to AWS Textract for Optical Character Recognition, which converts the image content into text (container number and size).
  4. Data Storage: The extracted data (container number, container size, date, and driver name) is stored in Supabase and associated with a unique ID for tracking.
  5. Image Storage: The highest-quality image detected by YOLO is uploaded to AWS S3 and linked to the database entry.
  6. User Interface: Users can access the main page, where they can view real-time detected container images, download videos, and review container data.

Contact and Feedback

If you have any questions, ideas, or feedback about the project, feel free to reach out to me via email at:
songjeongjun320@gmail.com

I would love to hear your thoughts and discuss any potential improvements or collaborations!

Demo Videos

Demo video: 1 mins Demo video: 2 mins