Geopipe

Geopipe is a New York City-based technology company that applies artificial intelligence to raw sensor data — aerial and drone photography, LiDAR and other laser scans, orthoimagery and point clouds — to automatically reconstruct cities as interactive 3D digital twins. Its proprietary machine learning pipeline identifies individual objects within a city and rebuilds them at multiple levels of detail, with regions reconstructed in advance so customers can download models immediately rather than commissioning manual 3D modeling.

The resulting environments are delivered for use in game development, simulation, training, architecture/AEC visualization and VR. Customers can stream real-world cities into their own software via a Unity SDK, 3D Tiles support and a RESTful API, work in Unreal Engine, or download static meshes in formats including FBX, OBJ, DAE, GLTF/GLB and OSG. Pre-built showcase environments include New York City, Boston and Washington, DC. Through a "Locations on Demand" service, Geopipe will build a digital twin of a requested region either from customer-supplied data (minimum 20 cm or better orthoimagery plus a point cloud at 0.7 m ground point spacing or better, or a photogrammetric mesh from which those inputs can be extracted) or from data sourced through its own data partners. The company describes its long-term goal as an authoritative whole-Earth digital twin built by AI.

Founding story

Geopipe was founded in 2016 by Dr. Christopher Mitchell and Dr. Thomas Dickerson. Mitchell, who holds a Ph.D. in computer science from NYU focused on distributed systems along with an M.S. in computer science and M.E. and B.E. degrees in electrical engineering, traces his interest in immersive copies of the real world to an attempt to recreate New York City in Minecraft. Dickerson completed his Ph.D. in computer science at Brown University, with a B.S. in computer science, mathematics and physics, and research applying the algorithmic lens to interdisciplinary problems; his interest in semantic modeling of built environments grew from wanting to automatically simplify models of real-world landmarks into virtual toy bricks while retaining their iconic features.

Business model

Geopipe licenses AI-generated 3D environments to enterprise customers, either through a SaaS API that streams cities into customer applications or as direct downloads of 3D models in standard interchange formats. It also performs bespoke reconstruction work, processing customer-supplied sensor data or partner-sourced data to build digital twins of specific requested locations.

Enterprise licensing of digital twin models via a SaaS API or direct model downloads, alongside on-demand reconstruction of customer-specified locations. Non-dilutive government research funding has also supported the company, including NSF SBIR Phase I and Phase II awards.

Traction

Geopipe received a $224,000 NSF SBIR Phase I award and, in May 2019, a $750,000 NSF SBIR Phase II award to commercialize its technology, which the company said enabled creation of two full-time and three part-time jobs and retention of three others. It was named 2019 SBIR/STTR Recipient of the Year by the New York district office of the US Small Business Administration. Following its seed round, the company said it planned to double the level of detail of its digital twins, accelerate coverage expansion, and expand deployments across gaming, simulation and AEC over the subsequent 12 months while hiring in research, engineering, product and business development.

Latest developments

Geopipe announced a $2.4 million seed round led by Village Global, joined by Matrix Partners, the Alexa Fund, AME Cloud Ventures and others, earmarked for deploying additional deep learning techniques, doubling digital twin level of detail, expanding geographic coverage, growing deployments in gaming, simulation and AEC, and hiring across research, engineering, product and business development. The company has also added Unreal Engine support alongside its existing Unity SDK, 3D Tiles and API offerings.