One of the highlights of Adobe’s annual MAX conference is its so-called ‘Sneaks’ session, where the company showcases some of its most groundbreaking research projects. Some of these make it the most important Creative Cloud apps, and some simply remain cool demos. At this year’s MAX, one of the most interesting sneaks is Project Strike a Pose.
Here’s the problem it’s trying to solve: sometimes you may have a picture of a model you want to use, but you do not have a picture of that specific model in the exact pose you need. Project Strike a Pose then lets you use another photo of another model in the correct pose as an example, and the software – powered by the Adobe Sensei AI platform – will automatically generate a new image with the original model in the same pose. It’s basically a kind of style transfer – but for model positions.
Image credit: Adobe
What is impressive here is that the result is not just a rough approximation of the sample pose or just the faces that have been switched between the source and the example. At least in the demo, Adobe’s Krishna Kumar Singh, the researcher behind the project, showed me that Project Strike a Pose’s neural network was able to invent the right look for the clothes the model is wearing, for example, and even rotate the model’s head after needs – and it even gets the shoes right.
Impressively enough, this even works when you ask the tool to imagine a picture of the model’s back – although as you can see in the example below, the hair here is a bit off.
Image credit: Adobe
Adobe would not say too much about how it trained the algorithms behind this project. However, it is clear that in order to train a neural network like this, it would need to use a lot of examples of models in a wide variety of poses. Singh has previously worked on a lot of research related to generative adversarial networks, so chances are that this new project is already based on this technique.
Right now, this is just an experimental research project, and as with so many of the company’s sneaks on MAX, it may or may never become a tool like Photoshop. However, if it works as well as it does in the demos (and Adobe says that the software also has no problems handling models of different ethnicities), it would definitely be a feature that many users would like to see in something like Photoshop or even in a tool like Adobe Spark that is already targeted at a wider audience with fewer image manipulation capabilities.
Unfortunately, it is also easy to see how someone could abuse a system like this. In an age of deep forgery and manipulated photos, this could be an easy way for someone who is not technical to create a compromising image of a public figure – or anyone else for that matter. Of course you can already do that today, but it takes a lot of skill to do it. Project Strike a Pose would make it a matter of a few clicks.