Tuesday 16:35–17:05 in Track 2

Step by step face swap.

Sylwester Brzęczkowski

Audience level:
Novice

Description

I will present how to implement a face swap. We will go step by step from simple “copy & paste” face on another image to fully functioning and nice looking face swap.

Abstract

Some people think Computer Vision is magical... well it's not. It's just math, sweat and software engineering. In this talk, I will present how to implement a face swap. We will go step by step from simple “copy & paste” face on another image to fully functioning and nice looking face swap. Everything is backed with live examples right there on the stage. The talk will give you a general understanding of what kind of operations are performed under the hood of face swapping applications. You will know how these operations work and how exactly they improve the solution. The talk will resolve the following problems:

  1. Face detection.
  2. Facial landmarks detection.
  3. Simulation of nonlinear image transformation using triangles warps.
  4. Finding such triangles.
  5. Blending 2 images together.
  6. Video stabilization.

Subscribe to Receive PyData Updates

Subscribe