Pittsburgh-based computer vision specialist Hellbender is preparing to launch a Raspberry Pi-powered stereo vision camera featuring a Hailo-8 accelerator for on-device machine learning and artificial intelligence (ML and AI) workloads.
“Hellbender On-Edge AI Stereo Camera is a fully programmable, industrial-grade depth perception sensor designed to revolutionize how robots see and navigate their world,” the company claims of its creation. “Built for ground mobile systems, field robotics, and autonomous platforms, it delivers professional-grade machine vision capabilities in a compact, power-efficient package that processes everything locally — no cloud required.”
The camera is built around the Raspberry Pi Compute Module 4 (CM4) computer-on-module — not the more recent and considerably more powerful Raspberry Pi Compute Module 5 (CM5). That may not matter, however, as Hellbender has chosen to pair the board with a Hailo-8 coprocessor that brings up to 26 tera-operations per second (TOPS) of minimum-precision compute to bear on heavier machine learning and artificial intelligence workloads. This is on top of the Raspberry Pi CM4’s quad-core Arm Cortex-A72 processor, running at up to 1.5GHz, and Broadcom Videocore-VI graphics processor.
The compute subsystem is connected to a pair of two-megapixel global-shutter full-color, arranged for stereo imaging with a 140mm baseline and a 129×78° field of view using the stock lenses. There’s an integrated 940nm structure light pattern projector with a 78×84° projection field plus an 850nm infrared flood illuminator with 104×84° coverage. An integrated six-axis inertial measurement unit (IMU) is included for visual-inertial navigation systems (VINS), alongside temperature, pressure, and humidity sensors plus a real-time clock (RTC) with battery backup.
The vision module is designed to be compact yet robust, and includes fins for passive cooling. (📷: Hellbender)
“For developers, we’ve made integration as simple as working with a Raspberry Pi,” the company claims. “Unlike development boards that require extensive additional components, Hellbender On-Edge AI Stereo Camera arrives as a complete, tested system. Simply connect to a compatible PoE [Power-over-Ethernet] injector or switch, and you’re ready to start developing.”
More information is available on the Hellbender website; the company is planning to launch the device on Crowd Supply in the near future, with interested parties asked to sign up to be notified when the crowdfunding campaign goes live.
