Looking Through AI-Tinted Glasses

The past decade has unquestionably seen fantastic advances in both the algorithms and hardware needed to build powerful artificial intelligence (AI) applications. But many of the applications — while they are very impressive in many ways — are still leaving a lot of people scratching their heads and wondering what exactly they can use them for. After you have generated a few pictures of a cat in a spacesuit riding a horse on Mars, and worked up a sweat trying to trick a large language model into saying some naughty things it is not supposed to, what is left?

One of the more interesting use cases, for generative AI (GenAI) in particular, is in assisting with creative pursuits. A mechanical engineer by the name of Oscar Wilmerding has put this strength of GenAI to work in building a very cool project called AIeye. It is a new take on a digital camera that snaps a picture, then uses AI algorithms to describe what is happening in the scene in detail before generating a reimagined version of it in a new, user-selectable style. Have you ever wondered what your backyard would look like if it were laid down on canvas by a post-impressionist? That is exactly the sort of thing that AIeye can show you.

The camera itself isn’t exactly a work of art. It was 3D-printed, so rough edges, chunky features, and layer lines abound. It follows the basic form factor of most digital cameras, with a display screen, scroll wheel, and a few buttons on the back. Inside, the device is powered by a Raspberry Pi Zero 2W single-board computer, and a custom PCB that ties together all of the supporting components. All in, the parts will cost about $100, so it is pretty reasonable for an AI-powered camera.

What really sets the AIeye apart from traditional cameras happens after the button is clicked to take a picture. The captured image is first passed to an online service called astica Vision. This service analyzes the image then returns a caption that explains what is present in the scene with a lot of detail. Next, the caption is sent to OpenAI’s DALL-E 3 text-to-image generator. DALL-E 3 then returns an image representing the caption in a style chosen by the user.

So what is AIeye useful for? Don’t you want a camera to actually take a picture of what you point it at? Well, normally you do. But as a creative aid, this camera looks amazing. It has the potential to help you see something in an entirely new way. Aside from that, it also looks like a lot of fun to play with, so it could be a great toy as well.

The parts list, 3D models, software, and build instructions are all available on GitHub. If you would rather buy it all together as a kit, be sure to fill out the form to express your interest.