Introducing JeVois-Pro deep learning camera

Greetings -

There has been quite some interest on this forum in our JeVois smart machine vision camera. Today we are launching a new, more powerful version on Kickstarter. Please check it out if you are interested!

As previously, we are very interested in helping teams develop their own vision pipelines. Please reach out to [email protected] or post comments on the Kickstarter page if you have any questions.

Here are quick specs:

  • CPU: Amlogic A311D with 4x A73 @ 2.2 GHz + 2x A53 @ 1.8 GHz.
  • GPU: Quad-core MALI G52 MP4 @ 800 MHz.
  • NPU: 5-TOPS integrated Neural Processing Unit.
  • RAM: 4 GB LPDDR4-3200.
  • Camera: 2MP Sony IMX290 back-illuminated Starvis sensor, 1/2.8”, 12mm lens, 1920x1080 at up to 120fps, rolling shutter, wide dynamic range support.
  • IMU: TDK InvenSense ICM-20948 with 3-axis accelerometer, 3-axis gyro, 3-axis compass, SPI bus @ 7 MHz, can be synchronized with camera sensor.
  • HDMI 2.1 video + sound output, up to 4k @ 60 Hz.
  • MicroSD card slot, up to 104 MByte/s, for operating system, software and data.
  • 2x USB 2.0 Type A ports (for keyboard, mouse, wifi, ethernet, etc).
  • 1x mini-USB OTG port.
  • 4-Pin UART (serial) port.
  • 6-pin auxiliary power out for 5V, 3.3V, and 1.8V peripherals.
  • 8-pin GPIO port (I2C + SPI, or 6x GPIO + GND + I/O voltage select).
  • Custom camera sensor connector, supports 1 or 2 sensors, 4x MIPI-CSI + IMU.
  • M.2 E-Key slot for 2230 PCIe x1/USB/SDIO/PCM/UART add-on cards (Coral TPU, WiFi, etc), supports custom JeVois extension for eMMC flash. (Note: PCIe x2 not supported).
  • Single 6-24 VDC input, 30 Watts max (including up to 15 Watts to power USB peripherals). Idle: 3 Watts. Running YOLOv2 on NPU: 5.3 Watts. Running CPU+NPU+TPU+VPU quad YOLO/SSD demo shown in video: 12 Watts.
  • Ubuntu 20.04 LTS (long-term support) aarch64 full.
  • OpenCV (latest) + OpenVino + all contribs and Python bindings preinstalled.
  • JeVois Core library with 30+ included machine vision modules preinstalled.
  • OpenGL ES 3.2, Vulkan 1.0, OpenCL 2.0, Coral Edge-TPU libraries.
  • Python 3.8 + numpy + scipy pre-installed.
  • Boost, Eigen, ImGui, glm, and many other C++ libraries pre-installed.
  • Install any extra aarch64 Ubuntu or Python packages using apt-get and pip3.
  • TensorFlow-Lite 2.5, Caffe, ONNX, MxNet, and Darknet deep learning support.
  • Import your own custom deep learning models to run inside JeVois-Pro.
  • Program your own machine vision pipelines in C++ or Python.
  • Full cross-compilation environment allows you to first develop and test your code on a standard Ubuntu Linux PC host computer, then cross-compile the same code for execution on the JeVois-Pro camera.
  • Weight: 80 grams (2.8 oz) with case, fan, heatsinks. Electronics only: 40 grams (1.4 oz).

best regards and keep up the great work!

– laurent


How can we get you more involved with FIRST to put these in the kits and help them understand that we need fiducials on the field?

Seriously, this looks awesome.


While I would not want to solve exact missions for the students, we can provide sample code that roughly points them in the right direction. A couple years back, we programmed these detectors to provide some inspiration:

And we would be happy to to the same on the new model. The barrier to entry should be significantly lower with this new camera, since it is a full-blown computer with ports for keyboard, mouse, display, etc so programming and debugging is much easier. Also you can customize your lens easily (standard M12 mount), including using lenses with no IR filter.


This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.