Our simulation platform generates synthetic data to train and improve machine learning algorithms.
The success of deep learning, a way to approach machine learning, has brought an insatiable hunger for data. The performance of deep learning often correlates with the amount of data used in training. However, data at scale is often proprietary, expensive, and laborious for people to manually prepare. The best way to deal with these challenges is with synthetic data — data created in virtual worlds rather than collected from the real world.
With synthetic data, a fully scalable data solution is now at our fingertips — at a fraction of the cost. By simulating the real world, virtual worlds create synthetic data that is as good as, and sometimes better than, real data.
We partner with businesses across industries to support their unique data needs.
We build photorealistic virtual worlds to closely mimic any real location where our client's services are being used.
Photorealism ensures that our synthetic data is effective in training AI to operate in the physical world. Because simulations are easier to control, these virtual worlds are the best place to test, train, and improve AI.
We offer diverse images and scenarios to help algorithms generalize well.
Photorealistic objects, difficult to reach places such as underwater locations, hard to replicate scenarios like extreme weather, varying perspectives from top down to bottoms up, costly real world tests like crashes all become easily simulated in virtual worlds.
We provide fully annotated synthetic data that can be used to immediately train AI.
Annotations such as 2D and 3D bounding boxes, semantic and instance level segmentations, surface normals, depth masks, edge masks, velocity annotations are all automatically generated in real time.
Synthetic data is a powerful method for training AI because of its scalability and flexibility.
UX+Data MEET UP
Fake It to Make It
Synthetic data promises massive sets of perfectly generated training data for a fraction of the cost of manually sourced annotated data. But there remains doubt about the efficacy of using synthetic data sets to train machine learning amongst practitioners. In this talk, Daeil Kim, CEO + founder of AI.Reverie reviews work of simulations for synthetic data in application verticals that are traditionally difficult to manually acquire significant data sets.
View the full version (45 minutes).
10 OCTOBER 2018
O’Reilly AI Conference
"AI.Reverie will be able to accelerate the performance of vision algorithms across many vertical markets, at scale and low cost and we are excited to be their partners."
Raanan Bar-Cohen, Co-Founder at Resolute Ventures