Artifical intelligence & Image generation.
GAN technology, based on Artificial Intelligence (AI), has enabled researchers to create realistic-looking but computer-generated photos of people’s faces. They have also allowed the creation of fictitious and controversial deepfake videos. In fact, GANs can be used to mimic any data distribution (image, text, sound, etc.).
Step by step
- Laptops with Python3, Anaconda, JupyterLab amongst other packages TBC
- A PC with a GPU would be useful but not essential.
- Google Drive access
Introduction to GAN technologies:
Introduction to GAN technologies. These technologies are explained step by step and with practical examples.
1.- Here is a rather terrifying application of GANs:
This is a GAN trained on thousands of celebrity faces to create unique faces that are scarily realistic!
2.- Here is a less terrifying video I made using a GAN, which I trained using 2000 images of “sunsets” from google images (that were “scraped”) This took maybe 2 hours to “train”.
This is made from the script I will be sharing, so participants will be able to make similar – but certainly much better/more inventive- videos themselves. Many different effects and levels of abstraction can be obtained by chnaging parameters of the GAN
3.- Here is another, more abstract, that was trained from the movie Song to Song (2017) by using frames as images: