Automating Access with Face Recognition

Automating Access with Face Recognition

How we built an AI-powered multi-location access control
system using advanced image and video analytics

As one of the fastest-growing technology and analytics companies in the United States, with multiple office locations, hundreds of employees, and 4000+ clients across the globe, Rapidops wanted to rethink the access control for our offices by replacing old-school access cards with AI-powered automated access control.

Key
Takeaways

  • Enacted Cognitive AI touchless access for attendance and seamlessly integrated access tracking.
  • Slashed costs by removing the need for cards while bolstering security against spoofing threats.
  • Completed the project within 6 weeks, ensuring effortless adoption and minimal disruptions.
  • Elevated employee experiences through direct walk-in access, automated attendance processes, and heightened security measures.
Power BI insights
Power BI insights
Power BI insights
Power BI insights
Power BI insights

Challenge

With organizations opening post-COVID and employees returning to offices, we found that accessing premises using ID cards posed a serious risk of spreading viruses and threatened employee safety. The ID card access systems also ran the risk of being tampered with, affecting login hours and, therefore, employee salaries.

As an innovation-first company, we wanted to reimagine access control by leveraging the latest technologies in AI/ML to build a centralized contactless identity verification system. Our vision was to deploy an advanced system that would be touchless, accurate, and smart to improve our employees' experience and the security of our office premises.

Solution

Rapidops’ engineering and innovation team conducted discovery sessions with our employees and HR teams to understand their issues with the current access systems. After learning about these challenges, we went to the ideation room to build a roadmap for the system. By mapping out the employee journey and identifying the key areas that could be improved to enhance their experience, we developed a customized face recognition engine and access control platform using Cognitive AI technologies.

Within six weeks, our team delivered a real-time face recognition engine and access control platform that enabled our employees to replace access ID cards with a hands-free experience, resulting in better employee satisfaction, operational efficiencies, and secure premises.

  1. Touchless Access

    By allowing for touchless entry and exit at all access points for all office locations, the system elevated the employee access experience. Employees no longer had to wear access cards or worry about losing or forgetting them. Designed with smart detection capabilities, the system could also detect multiple employees in a frame and faces in low light.

    Visitor & Delivery Management
    Visitor & Delivery Management
    Visitor & Delivery Management
    Visitor & Delivery Management
    Visitor & Delivery Management
  2. Automated Attendance

  3. Visitor & Delivery Management

  4. Centralized Operations

Powered by Deep Learning & Cognitive AI

The system was built with Cognitive AI technologies like real-time object detection, deep learning, and on-demand model training that allowed for higher accuracy of results. Our engineers used open-source technologies like OpenCV, MTCNN, and FaceNet to detect face edges from input images, extract features and train the model on these to recognize employees' faces.

Centralized Access Server
  • Built custom hardware using Rasberry Pi to control door operations
  • Built a real-time model training system with a live camera feed
  • Implemented YOLO models using OpenCV for detecting multiple faces in real time
  • Implemented Triple Loss Function from FaceNet to improve the accuracy of model detection
  • Implemented Support Vector Machine ML model for higher search accuracy and faster detection results
  • Applied Liveness Detection to ensure prevention of spoofing attacks using fake or non-real faces

How we
did it

Analytics & AI/ML

  • Real time data analysis
  • Predictive analytics
  • Deep learning
  • Neural Networks
  • 2D & 3D Mapping
  • Generative Adversarial Networks

Strategy

  • Product Strategy
  • Data Strategy
  • Analytics & AI/ML Strategy
  • Roadmapping

Technologies

  • PySpark
  • Solr
  • Cockroach
  • Docker
  • Kubernetes
  • Python
  • FastAPI
  • Spacy
  • NLTK
  • Gensim
  • Pandas
  • PyTorch
  • Requests

Engineering

  • Native mobile on IOS & Android
  • Web app development
  • Automation framework & manual QA
  • REST API backend
  • Front end development
  • ADA Accessible UI
  • IoT development with Raspberry Pi