DexForce's Xema Is a High-Accuracy Industrial Smart Depth Camera Powered by an NVIDIA Jetson Nano
Offering a claimed 50 micron accuracy, this smart depth camera has been released under the permissive Apache 2.0 license.
Computer vision start-up DexForce has released an industrial-grade 3D depth-sensing camera system with a claimed accuracy of 50 microns — and it's making it available under an open source license.
"The goal at current stage is to create a low cost 3D camera with capability of current generation industrial cameras," the company explains. "It is also a computer with GPU computing capability, which makes running 3D point cloud recognition algorithms and robotic arm control programs possible. This camera can be used in scenarios such as bin-picking, assembly, 3D inspection and many others. Our ultimate goal is to continuously reduce the cost of industrial 3D cameras with the help of deep-learning's capabilities of abstract representation, and allow machine vision and AI to penetrate all aspects of this carbon-based world."
Brought to our attention by Seeed Studio, the DexForce Xema is built around an NVIDIA Jetson Nano computer-on-module, giving it the capabilities it needs to handle 3D point-cloud generation and machine learning algorithms on-device as well as handle interfacing with external hardware. A CMOS camera sensor combined with a DLP projector give the device the ability to accurate capture depth information — while DExForce's "Mixed AI" software aims to make using that data as easy as possible.
To boost the ecosystem around the device, DexForce has chosen to make the core Xema design available under an open source license. As released, the camera offers a 0.05mm calibration accuracy, image and depth resolutions of 1920×1200, and a working range of between 400mm and 2,000mm (around 15.75" to 78.74"). Those looking for real-time operation, though, are advised that the camera captures at the rate of just one frame per second.
"All of our development is open source, that anyone can use and improve it," the company claims of the device's design ethos. "[And] a growing number of detailed tutorials will help getting started." At the time of writing, however, the majority of the project's documentation was available exclusively in Chinese.
More details on the project alongside source code and design files are available on the Xema GitHub repository under the permissive Apache 2.0 license.