How it works

We use artificial intelligence (AI) to generate design for objects. In a first application for the AI model we trained we generate new examples of necklaces.

The story behind necklaces

The idea behind the necklaces was to take an object that had been part of human history for a long time and at the same time not limited by shapes that had to fit - because of the uncertainty of the output.

The process of generating necklaces

We use generative adversarial networks (GANs) to generate design for necklaces, because it gives the best results when it comes to image and object generating tasks. The data consists primarily of images of necklaces and pendants. The models are trained on Graphical Processing Units (GPUs) because the learning procedure is compute intensive. The AI model generates output that is scripted into 3D-models, and is what we have used for the AI collection. Some manual fine tuning is needed, for example ensuring there is hole for the chain, because the model has difficulty learning all the features. The output of the AI model is then sent to a Swedish foundry that cast the necklaces.

The AI model generates output that is scripted into 3D-models, and is what we have used for the AI collection.

Some manual fine tuning is needed, for example ensuring there is hole for the chain, because the model has difficulty to learn all the features. The output of the AI model is then sent to a Swedish foundry that cast the necklaces.

The future path

We want to create an end-to-end solution where we can use generative models in 3D generation based on user input and preferences, in order to generate personalized and unique models, which are then 3D-printed and delivered to the user.

This site uses cookies to give you a better user experience. Read more I accept