Home

Awesome

Background Stylizer - with Segmentation & Style Transfer

This is the GitHub repo for an end-to-end tutorial from TFLite model conversion, to model deployment on an Android and iOS. This project received the #TFCommunitySpotlight award for its creativity and real-world user scenarios. This is one of the End-to-End TensorFlow Lite Tutorials. See the full list of TensorFlow Lite samples and learning resources on awesome-tflite.

Links to the multi-part tutorial blog posts on Medium:

  1. Project Intro, Margaret Maynard-Reid
  2. TensorFlow Lite models, Sayak Paul
  3. Android implementation, George Soloupis
  4. iOS implementation, Patrick Haralabidis (coming soon)

This project combines multiple CV models to create visually appealing demo that run in real time on device. For example, [segmentation plus style transfer]; or segmentation plus background removal (1 | 2)

<p align="center"> <img src="images/segmentation.png" width='680'/> <img src="images/style_transfer.gif" /> </p>

Demo created by Sayak Paul (Colab Notebook)

<p align="center"> <img src="images/sayak_demo.png" width='742'/> </p>

Demo created by Khanh LeViet (Colab Notebook)

<p align="center"> <img src="images/khanh_demo.png" width='742'/> </p>

Acknowledgement

We would like to thank the TensorFlow Lite team and the ML GDEs Google Developers Expert Program for their collaboration and continuous support. This project is also a great showcase of community collabration by members of diverse geographic areas (U.S., India, Greece and Austrialia), expertise and background.

<div align="center"><img src="images/made-by-ml-gdes.png" width='96' height='96'/></div>