Artificial intelligence is not only used to make predictive models in sectors such as medicine and weather, or facial recognition applications. If last week we saw how this use reached the generation of memes, now we see it mixing animal faces in humans of all kinds.
As collected on Motherboard, it has been Xander Steenbrugge, a researcher and educator of machine learning, who has decided to create a model that combines the faces of up to five types of animals with humans.
From random generation to inter-species mixing
This, in addition, has decided to publish the model used so that everyone can modify it and make their own “humanimals”, as you have called this curious mixture. This processing can be done continuously, such that the incremental steps that make the transition between some species and others are generated.
In this way, Steenbrugge even generates videos that transition from the individuals of each species – also generated by AI in the model itself – to others of another, in an animated and fluid way. At a technical level this is achieved through a type of neural network called GAN, or antagonistic generative network, from images also generated.
The researcher, who is recognized as passionate about artificial intelligence, has created these types of transitions between images previously. The most impressive are collected on his YouTube channel, Neural Synesthesia, and combine music with the transformation of people, objects and landscapes into others completely different, which is taken to the next level with the publication of this latest model.
Steenbrugge has released the details for this software. And not only that, but it is possible to generate the faces of animals in humans in a relatively simple way. As easy as executing the code directly in Google Colab, Google’s online notebooks, and ** getting new images – yes, at a relatively modest resolution – with just a few parameters.
👇 More in Hypertextual