Experiment: Aged-Up Deepfakes
From this TETRA research project AI in production, we have prepared an introductory workshop on deepfakes to give to the DAE students of the VFX major. During the annual creative week, all regular courses for first and second year students are canceled and replaced by creative sessions. Our goal was to provide an introduction to deepfakes from the research group: what is deepfake technology, what can you do with standard software without heavy programming yourself, what are the points of interest if you want to get started with it, …
Due to the lockdown, the creation week was canceled this academic year, but the preparations for the workshop were already ready. The workshop with students will be moved to a later date in one of the VFX classes, but we are happy to share the approach of the workshop with the guidance group here. Read about the limits of this technology and how we tackled Star Wars in a hands-on use case with the combination of different tools.
Reinforcement Learning to find bugs
In this overview post we show how Reinforcement Learning (RL) can be applied to testing games. We do this by means of several papers and use cases. We show how you can detect errors in level design and game breaking bugs in this way. It may be helpful to read our introduction to RL first if you are unfamiliar with the concept.
2D to 3D: Building environments using photoshop input
Often only 2D data is available from buildings, machines or environments that contain information about how something is constructed. When creating a 3D environment, it is also faster to draw a floor plan than to build an environment yourself in 3D. The purpose of this case was to explore the possibilities and workflow regarding the construction of a 3D environment based on 2D input.
Leaf material for 3d production
Simulating plant growth has had social and scientific relevance for a long time for various purposes, mainly for agriculture. This takes into account the biological research that shows the influence of the plant hormone auxin on growth. This knowledge has been converted into a mathematical model for the simulation of leaf growth. In addition, a growth function, which is influenced by auxin maxima on the leaf margin, determines the shape of the leaf. The literature describes techniques that are biologically accurate, but are not intuitive due to their complexity and are also time-intensive. Simplified models can be found, but often lack identifying characteristics, such as serrated edges around the blade and true-to-life veining. For this use case, the mathematical model of leaf growth and shapes from Runions et al. was used in combination with a simplified leaf shape.
Code synthesis with LLM
Good news for people without programming experience. You can now put together a game without even typing a letter code. We can now indeed automatically generate code with language models. This is done using simple commands, or the model simply supplements existing code. And for beginners: a language model can check your code for syntax errors and also optimize them automatically. Even for more experienced programmers, there is sometimes still rehearsal or research work. Especially when juggling between different languages. In this blog post we see how programming is now easier than ever before.
