Implementation of a gender recognition algorithm using Xception NN

Project by Jeremy

Abstract

In the ecosystem of products from iCarbonX, the Smart Mirror is a connected mirror to help people to do gym exercises, give precise measurements of the body and more… For a company product, it’s needed to detect the gender of the person just by using its picture. Project requirement is starting from scratch, build a model using Xception, RGB & Depth images.

 

Challenges

  • Lack of data / CNN architecture
  • Implement it on product

Achievements (according to KPIs)

  • High Resolution model
  • Increase the dataset using different flows ( youtube scraping & Face detector)
  • Model tested on Mirror -> Testing step before release

 

Further development

Make a lighter model to be more responsive, detect age to be careful on predictions on kids (challenging part)

Share this post

Share on facebook
Share on twitter
Share on linkedin
Share on email