Overview
This course provides an introduction to unpaired image-to-image translation using CycleGAN. The learning outcomes include understanding the concept of instance normalization and its application in maintaining mean activation close to 0 and standard deviation close to 1. The course teaches how to download instance normalization code and implement it in the context of image translation. The teaching method involves a walkthrough of key details from the original paper, discussing discriminator architecture (PatchGAN), generator description, generator architecture, generator training, and the combined/composite model. The intended audience for this course includes individuals interested in image processing, deep learning, and computer vision.
Syllabus
Unpaired image to image translation Using CycleGAN
Walkthrough - Key details from the original paper
Discriminator architecture (PatchGAN)
Generator description
Generator architecture
Generator training
Combined / composite model
Taught by
DigitalSreeni