Refresh

This website linuxgizmos.com/quad-camera-rig-taps-into-jetson-agx-xavier-for-deep-learning/ is currently offline. Cloudflare's Always Online™ shows a snapshot of this web page from the Internet Archive's Wayback Machine. To check for the live version, click Refresh.


All News | Boards | Chips | Devices | Software | LinuxDevices.com Archive | About | Contact | Subscribe
Follow LinuxGizmos:
Twitter Facebook Pinterest RSS feed
*   get email updates   *

Quad-camera rig taps into Jetson AGX Xavier for deep learning

Jul 15, 2019 — by Eric Brown 2,521 views

E-con has launched a Linux-driven, AI-enabled “SurveilsQUAD” camera system for the Nvidia Jetson AGX Xavier or TX2 with 4x 2-megapixel cameras with HD or FHD resolution connected via MIPI-CSI-2.

E-con Systems has begun shipping a SurveilsQUAD (e-CAM20_CUXVR) camera system with a V4L2 Linux driver and a sample Linux app with source. Like the robotics focused e-CAM130_CUXVR kit it launched in January, the SurveilsQUAD has four 4-lane MIPI-CSI-2 connected cameras and is designed to work with the Linux-powered Nvidia Jetson AGX Xavier Development Kit.

This time, however, the cameras are 2-megapixel instead of 13MP, and the application is surveillance enhanced by AI and deep learning applications running on the Xavier module’s high-end Nvidia Volta GPU. Applications include smart surveillance, parking lot management, intruder detection, and object tracking. (Although AI-enhanced surveillance kind of creeps us out, we realize the technology could also save lives in addition to enabling warrantless invasions of privacy.)



SurveilsQUAD camera board with cameras (left) and complete system attached to Jetson AGX Xavier dev kit
(click images to enlarge)

The SurveilsQUAD, which can also work with the developer kit for the lower-powered Jetson TX2, is built around E-con’s e-CAM21_CUMI290_MOD camera, which incorporates a Sony Starvis IMX290 color, 1/2.8” CMOS image sensor. The 2-megapixel camera supports HD (1280 x 720) and FHD (1920 x 1080) at 30fps in sync mode and either 120fps or 60fps in async mode, depending on whether you use 10-bit or 12-bit output. The system depends on the Xavier Image Signal Processor (ISP) to process Auto White Balance and Auto Exposure control functions.

— ADVERTISEMENT —


The S-Mount compatible system consists of the camera board and a baseboard to interface with the Xavier Developer Kit. The system supports -30 to 85°C temperatures and ships with 5x 30cm micro-coaxial cables and an SD card with the Linux code.

Nvidia’s 105 x 87 x 16mm Jetson AGX Xavier module has greater than 10x the energy efficiency and more than 20x the performance of the Jetson TX2, claims Nvidia. The module features 8x ARMv8.2 cores and a 512-core Nvidia Volta GPU with 64 tensor cores with 2x Nvidia Deep Learning Accelerator (DLA) engines. There’s also a 7-way VLIW vision chip, as well as 16GB 256-bit LPDDR4 RAM and 32GB eMMC 5.1. The module is available via a $1,300 Xavier Development Kit.

 
Further information

The SurveilsQUAD (e-CAM20_CUXVR) camera system is available in single-unit evaluation samples through July 23 for $549 plus shipping, after which the price will rise to $599. More information may be found on E-con Systems’ SurveilsQUAD product and shopping page.
 

(advertise here)


Print Friendly, PDF & Email
PLEASE COMMENT BELOW

2 responses to “Quad-camera rig taps into Jetson AGX Xavier for deep learning”

  1. Dmitry Kargin says:

    Correction: not `1/28”`, but `1/2.8” CMOS image sensor`.

  2. Jeff Child says:

    Thanks Dmitry.
    We’ve made the correction.

Please comment here...