At Speech Graphics, we invest heavily in R&D to push our procedural facial animation technology to new quality levels, and to engineer new solutions integrating this technology into future products. We have something groundbreaking in the works right now, but can’t say too much yet.
What we can talk about is one of our key research methods: deep learning. You might have heard some buzz around deep learning in the news. For the uninitiated, it is a method of machine learning that recognizes patterns in data. It has some similarities to how the audio-visual cortex in the brain processes stimuli.
Joris, our lead researcher in machine learning, gives an overview:
“Deep learning is not a new idea. It was developed many years ago, but has only gained widespread popularity in academia and industry in the last decade or so. This is due to improved ways to overcome its inherent difficulties, large computational resources becoming widely available, and most importantly, the increased performance it has demonstrated in virtually every task it is applied to compared to other techniques.
It is now widely used in many pattern recognition tasks, from things like automatic recognition of speech or images, to tasks like searching the web or predicting the stock market. Here at Speech Graphics, we see deep learning as a powerful additional tool to link facial motions with various kinds of related data, such as speech, text, emotion, etc.”
We are not the only ones who have adopted deep learning – Google, Microsoft and IBM have all recently announced major developments using deep learning as well. Deep learning will enable us to continuously improve our technology, and get closer than ever before to bringing next-gen to life.
We will share more information about our new animation projects and research findings soon. In the meantime, to find out more about Carnival™, the system driving Speech Graphics animations, check out our Technology page. To see what Speech Graphics can add to your next project, drop us a line.