DevMesh

Running real-time style trasnfer on a Raspberry Pi (4B-1GB)

Sayak Paul

Sayak Paul

Kolkata, West Bengal

0 0
  • 0 Collaborators

The project uses a light-weight Style Transfer model and runs that on real-time video feeds. The whole project is executed on a Raspberry Pi to enable to use of deep learning on embedded devices. ...learn more

Project status: Under Development

Artificial Intelligence

Intel Technologies
Intel Python

Overview / Usage

The project showcases the use of light-weight TensorFlow Lite models on a Raspberry Pi computer. It runs a style-transfer model on real-time video feeds.

Methodology / Approach

  • Grabbed the official TFLite models on Style Transfer. We need to use two models here - one for extracting the style vectors and another one for applying that style vector to stylize a given image.

  • Ported those two models to Raspberry Pi and used PyImageSearch's highly parallel `imutils` to handle the live video frames and apply the TFLite models for running inference in parallel.

Technologies Used

  • TFLite

  • OpenCV

Comments (0)