Sketch-to-Car: Synthesizing Stylized Car Images From Sketches

Code

• github.com/sugia/sketch-to-car
  In papernone★ 2

Abstract

We investigate the inception module as a building block of conditional generative adversarial network (GAN) in training a sketch-to-car model. We demonstrate that this approach is effective at synthesizing photos from edge maps. Experimental results show that, the use of inception modules significantly accelerates the training process of conditional GAN compared to those using convolutional layers, while maintaining a similar frechet inception distance (FID) score.

Tasks

Reproductions