CopernicAI

Generative AI Environments

Building the next wave of open generative AI for environments

Generative AI has undergone remarkable progress in recent years, with the quality of AI-generated text and imagery today continuing to amaze. Other data modalities are following. The most ambitious, perhaps, is to fully generate 3D environments.

The first step to generative environments is what we call a 2+1D approach, where generated 2D imagery is augmented with 1D depth information. Leveraging our recent research, published in the premiere deep learning conferences, we can realize generative 2+1D environments today, as we work towards full AI generated 3D environments.

AI generated 2+1D environments

Explore CopernicAI environments in VR

For the Quest headset (contact us for other headsets).

AI generated panoramic images

Roadmap

Efficient Panoramic AI

Published in ICLR 2021

Efficient techniques providing significant improvements in computational complexity and state-of-the-art performance on low-resolution tasks. But still limited to very low-resolution.
Partially Scalable Panoramic AI

Published in ICLR 2022

High-resolution input data but only low-resolution output data. Applicable for classical AI tasks with low resolution outputs (e.g. image classification, object detection and tracking).
Scalable Panoramic AI

Published in ICLR 2023

High-resolution input and output data. Achieves state-of-the-art for dense-prediction tasks (e.g. semantic segmentation, depth estimation). Building blocks for Generative Panoramic AI.
Generative Panoramic AI

Under Development

Generative AI panoramic images and videos using Scalable Panoramic AI building blocks. Coming soon…
Generative AI Environments

Next!

Generative AI for 2+1D (panoramic + depth) and full 3D environments.