Inquiry about Product for Object detection

• MartyG

  • March 05, 2020 16:16
  • Edited

  Hi Olivier,

  The Neural Compute Stick 2 can plug into a USB port alongside a RealSense camera and accelerate computer vision applications such as stream aligning so that it runs faster.  It is not a computer board though and so you would still need a computing device to attach the camera to and run it with an operating system such as Windows or Linux.

  There are small and affordable boards such as the Raspberry Pi 4 that are suitable for running a single camera. 

  https://www.raspberrypi.org/products/raspberry-pi-4-model-b/ 

  The RealSense cameras do not have a 45 FPS mode though - the supported FPS speeds either side of that value are 30 and 60 FPS.  So you would likely have to run the application in 60 FPS mode.  

  The Raspberry Pi 4 claims performance comparable with a desktop PC (more likely a business PC than a gaming PC).  If you are only using one camera, it is conceivable that the Pi 4 could cope with 60 FPS, and its price is well suited to a student project.  You could later optionally add a Neural Compute Stick 2 alongside the camera if you find that it is not running quite fast enough for your satisfaction.

  Intel's version of the OpenVINO toolkit is a free download and is available for both Windows and Linux.

  WINDOWS

  https://software.intel.com/en-us/openvino-toolkit/choose-download/free-download-windows 

  LINUX

  https://software.intel.com/en-us/openvino-toolkit/choose-download/free-download-linux 

  Regarding camera choice, for autonomous vehicles the D435 is the appropriate choice, as it has a fast shutter able to process data when a vehicle is moving at full speed (either observing a vehicle or travelling whilst mounted on one).  For a small amount more, you can get the D435i model that has a built-in tracking IMU with gyro and accelerometer.

  There are some Deep Neural Network (DNN) object recognition sample programs already available for RealSense.

  https://github.com/IntelRealSense/librealsense/tree/master/wrappers/opencv/dnn 

  https://github.com/IntelRealSense/librealsense/tree/master/wrappers/openvino/dnn

  https://github.com/twMr7/rscvdnn 

  0

  Comment actions Permalink
• Olivier Berton

  • March 06, 2020 09:23

  Hi,

   

  Thanks a lot for your great response.

   

  Sincerely,

  0

  Comment actions Permalink
• MartyG

  • March 06, 2020 10:15

  You are very welcome.   Please feel free to ask any follow-up questions when you need to.

  You may find the YouTube video in the link below of a D435 mounted to a vehicle moving at full speed to be a useful reference four your project.

  https://www.youtube.com/watch?v=OwJmCyAn3JQ 

  0

  Comment actions Permalink

Please sign in to leave a comment.