Realtime Currency detection with Seeing AI

AI is not just for data scientists. Developers can build practical applications that ride on the underlying machine learning models to build practical applications. In this session, we take a look at Seeing AI a deep neural network model for the visually challenged. We look at how we can integrate Seeing AI into applications and use it for currency detection.

With minimal slides and a more hands-on approach, we walk through the process of how developers can build, train and integrate the ML model into their applications and deploy the same on the EDGE for real-time detection of something as diverse as currencies.


Outline/Structure of the Talk

  • Introduction
  • Data preparation and pre-processing
  • Choosing the right model
  • Build and Train the model
  • Deploy the model to the EDGE
  • Currency detection and other practical examples
  • Summary

Learning Outcome

You can understand how to build and deploy a machine learning model integrated with an app on the mobile. The EDGE deployment allows models to be used even with no connection to the cloud and even without a internet connection.

Target Audience

AI/ML Developers

Prerequisites for Attendees

Python, TensorFlow, iOS/Android development.

schedule Submitted 5 months ago

Public Feedback

comment Suggest improvements to the Speaker
  • Dr. Vikas Agrawal
    By Dr. Vikas Agrawal  ~  4 months ago
    reply Reply

    Dear Praveen: Would you like to share a video of you presenting or introducing the topic?

    I wonder about currency detection applications have existed on ATMs and Vending Machines since the 1970s, and those were updated in the 80s, 90s and 00s, with a combination of detecting size, paper thickness, paper type, graphics elements, security elements etc. What are you doing fundamentally different with the solution you will show at ODSC? Can it be confused by a printout or image of the currency even without physical security features?

    Warm Regards



    • Praveen Srivatsa
      By Praveen Srivatsa  ~  1 month ago
      reply Reply

      We have a trained data set of over 15 currencies and coins. We use this for identification with even partial images. The demo is more to bulk identify (find the total amount in a wallet) amounts and does not check if the currency is legal. It can also work with video and can identify the total amount across currencies in a short clip.

      (Apologies for the late comment. Had some medical issues in the immediate family. Also I got the email for the session rejection, but did not get an email for the comments - so did not know about it before. Look forward to connecting next year)

  • Anoop Kulkarni
    By Anoop Kulkarni  ~  4 months ago
    reply Reply

    Thanks for your proposal. I have similar comments for this, like your other proposal. Do share which datasets you plan to use. Thanks