e-con Systems and Qualcomm launch ready-to-deploy edge AI vision kit

e-con Systems and Qualcomm launch ready-to-deploy edge AI vision kit

Apr 13, 2022 — by Giorgio Mendoza

69 views

Early this month, the strategic partnership between e-con Systems and Qualcomm launched a ready-to-deploy AI vision kit targeted for applications that require powerful vision computing, machine learning and wireless connectivity. The edge AI vision kit consists of the e-con’s Sony STARVIS IMX415 and the Qualcomm QCS610 processor.

The qSmartAI80_CUQ610 vision kit incorporates the Qualcomm® QCS610 System on Chip designed to deliver high performance without sacrificing thermal efficiency. The QCS610 SoC integrates a powerful image signal processor (ISP), a Qualcomm-based Artificial Intelligence Engine and customizable CPU, GPU and DSP to meet AI performance requirements.

Evaluation Kit Contents 
(click images to enlarge)

There are two Kryo Gold CPUs (up to 2.2GHz) and a Hexa low power Kryo Silver CPU (up to 1.8GHz) embedded in the QCS10 System on Module. In addition, there is a dedicated Qualcomm Snapdragon neural processing engine hardware which runs a deep neural networking algorithm optimized for machine learning and vision computing applications.

QCS610 SOM Block Diagram
(click images to enlarge)

On the vision side, the evaluation kit includes the e-CAM83_CUMI415_MOD_H02R1 which is a manual focus camera that was built to perform well in low-light settings. The camera module is enabled to support 4K at 30 frames per second (max performance). The camera uses CMOS technology (STARVIS) from SONY, to be more specific, is based on the IMX415 sensor. The IMX415 comes in a 1/2.8″ optical form-factor and it features a variable electronic shutter. 

QCS610 SOM on carrier board (left) and e-CAM83 camera module (right)
(click images to enlarge)

To accelerate software development e-con Systems provides GStreamer applications that involve, HDMI streaming (NV12), Network streaming over Ethernet and WIFI (H.264 encoded), Snapshot image capture in JPEG and RAW Bayer,H.264 encoded video recording and modifying camera features like White balance, Gain, Exposure etc. In regards to the AI development e-con Systems provides a Qualcomm-developed neural processing SDK, Neural networks APIs, GStreamer plugins(DLC models) and a GStreamer TFLite plugin. Refer to the Software section on the e-con Systems site. 

Specifications listed for the qSmartAI80_CUQ610 – Qualcomm® Edge AI Vision Kit include: 

• Processor:
  • Qualcomm® QCS610/410 SoC (QCS610 default processor)
    • Qualcomm® Kryo™ 460 CPU
      • Kryo Gold: Dual high-Performance cores 2.2 GHz
      • Kryo Silver: Hexa low-power cores 1.8 GHz
    • Qualcomm® Adreno™ GPU 612 with 64-Bit Addressing; 845 MHz
• Memory/storage:
  • 4GB LPDDR4X SDRAM, (up to 1866 MHz)
  • 16GB eMMC
• Networking 
  • Wi-Fi 2.4GHz/5Ghz, Bluetooth, ethernet
  • RF Interface: 1 x U.FL, 50 ohm coaxial connectors
• Interfaces:
  • Three 4-lane MIPI CSI for camera interfaces
  • MIPI DSI One 4-lane 
  • MIPI DSI display interfaces
  • 1x USB3.1
  • 1x USB2.0 (Type C) 
  • 1x USB 2.0 (Host only)
  • One 4-bit Secure digital interface (SDIO)
• Expansion:
  • 4x GPI
  • 4x GPO
• Other features – TPM 2.0 onboard IC and Watchdog timer (256 segments)
• Power – 12V DC
• Operating temperature — 0 to 60°C
• Dimensions:
  • 55 mm x 35.5 mm (SOM)
  • 30 mm X 30 mm (Camera board)

Relevant key features of the camera module:

• Sensor:
  •  Sony STARVIS® IMX415
• Focus Type:
  • Fixed focus
• Resolution:
  • 4K UHD (3840 x 2160)
• Output Format:
  • 10bit RAW Bayer
• Frame Rate:
  • 4K @30fps
• Optical Format:
  • 1/2.8″
• Shutter Type:
  • Electronic rolling shutter
• Pixel size:
  • 1.45 µm x1.45 µm
• Holder:
  • M12 (S-Mount)
• DFOV:
  • 115.90° (with the lens provided by e-con)

Further information

e-con Systems also offer customization support for the camera module, carrier board, enclosure as well as software customization (Custom BSP, ML and DL algorithms). Refer to the e-con Systems website for additional information about pricing and availability.