Mission

Enabling autonomy through robotic vision you can trust.

Robots fail when their cameras fail. The 3D information they need to manipulate and navigate the environment is too often noisy, incomplete, and unreliable. Or, just simply too expensive.

Level-5 autonomy, safe household robots, and dependable autonomous drones require vision that works everywhere; robustly, reliably, and like humans. The path forward is the same one humans use: a data-driven approach to perception.

Efference is redefining robotic vision by building the most intelligent cameras on the market.

Model

Seeing the world in 3D

Any two-dimensional image is consistent with an infinite number of possible 3D worlds. Determining which world is most likely is the fundamental problem of perception.

Our models solve this the way humans do, by selecting the most probable world using every available cue: shadows, perspective, occlusions, and much more. The result is clear, consistent 3D structure that you can actually train robots on.

Learn more about perception →

Camera

H-01

The H-01 is a passive stereo system modeled on the constraints of human vision. Two global-shutter 5 MP sensors capture high-speed RGBD video — 2560×1440 at 30 fps, 1080p at 60 fps, and 720p at 120 fps.

A 60 mm baseline, matching human interpupillary distance, is integrated into a 100 × 30 × 30 mm enclosure. A 140° diagonal FOV and HDR imaging provide wide coverage and reliable performance in real-world lighting.

To maintain stable depth during motion, the system incorporates dual 400 Hz IMUs that fuse inertial and visual information with extremely low latency. An automotive-grade GMSL2 interface enables long-distance, low-noise transmission to onboard compute.

The design philosophy is simple: tasks performed by humans are optimized for the human visual system. H-01 brings that same geometry and sensing strategy to robotics, with state-of-the-art algorithms running directly on the PCB board.

Team

Efference is built by biologists, neuroscientists, and computer scientists who have worked on the first manned mission to Mars, visual perception and decision-making in primates, and uploading the first biological creature, a fruit fly, into a computer simulation.

We are proudly supported by Y Combinator and Anti Fund.

We help robots see like humans do