Usepercept provides a foundation model that converts raw video, LiDAR, or satellite data into sub‑centimeter accurate, metrically referenced 3D scene graphs in minutes.
Funding
Funding not disclosed
Founders
Product
Problem
Engineers and operators often receive raw sensor feeds—such as video, LiDAR, or satellite imagery—that require extensive manual processing to extract usable 3D models, making it difficult to quickly assess objects, defects, and spatial relationships for planning or simulation tasks.
Solution
Percept offers a foundation model that automatically transforms raw sensor data into a metrically accurate, queryable 3D scene graph within minutes. The pipeline reconstructs the environment, enriches it with semantic labels, material properties, and defect annotations, and organizes the information as a structured graph that can be inspected visually or accessed programmatically. This representation supports physics‑based simulations—such as fire spread, structural failure, or flood modeling—directly on the real-world geometry, enabling both operators and developers to reason about the physical world without building custom perception pipelines.
Target Audience
Primary customers include public safety agencies, utilities, construction firms, and transportation planners using the visual interface, as well as drone operators, robotics developers, autonomous system builders, and edge‑AI teams leveraging the API.
Features
- End‑to‑end processing of video, LiDAR, or satellite imagery into sub‑centimeter accurate 3D models in real‑world coordinates
- Structured scene graph encoding objects, materials, defects, spatial relationships, and temporal state for direct querying
- Visual operating environment (SpatialOS) for non‑technical users to upload data, run queries, and launch simulations without code
- Programmatic API (Percept SDK) for developers to integrate reconstruction, querying, and simulation into autonomous, edge‑AI, and robotics applications
- Built‑in physics layer that enables realistic simulations of fire, structural failure, flood inundation, and other phenomena on the generated geometry
- Semantic labeling and adjacency/occlusion indexing to support advanced reasoning and risk‑zone identification