Home > Broadcom BCMxxxx, Debian, Linux, Video > GPU Accelerated Object Recognition on Raspberry Pi 3 & Raspberry Pi Zero

GPU Accelerated Object Recognition on Raspberry Pi 3 & Raspberry Pi Zero

You’ve probably already seen one or more object recognition demos, where a system equipped with a camera detects the type of object using deep learning algorithms either locally or in the cloud. It’s for example used in autonomous cars to detect pedestrian, pets, other cars and so on. Kochi Nakamura and his team have developed software based on GoogleNet deep neural network with a a 1000-class image classification model running on Raspberry Pi Zero and Raspberry Pi 3 and leveraging the VideoCore IV GPU found in Broadcom BCM283x processor in order to detect objects faster than with the CPU, more exactly about 3 times faster than using the four Cortex A53 cores in RPi 3.

They just connected a battery, a display, and the official Raspberry Pi camera to the Raspberry Pi boards to be able to recognize various objects and animals.

The first demo is with Raspberry Pi Zero.

and the second demo is on the Raspberry Pi 3 board using a better display.

Source code? Not yet, but he is thinking about it, and when/if it is released it will probably be found on his github account, where there is already py-videocore Python library for GPGPU on Raspberry Pi, which was very likely used in the demos above. They may also have used TensorFlow image recognition tutorials as a starting point, and/or instructions to install Tensorflow on Raspberry Pi.

If you are interested in Deep Learning, there’s a good list of resources with links to research papers, software framework & applications, tutorials, etc… on Github’s .

  1. crashoverride
    April 30th, 2017 at 14:58 | #1

    “Source code? Not yet, but he is thinking about it”

    That should have been in the headline. If there is no source code, then its just a hoax! 😉

    • Pablo Prats
      May 1st, 2017 at 03:28 | #2

      No source code is a hox

  2. Mic_s
    April 30th, 2017 at 22:04 | #3

    No, it’s not a fake. It’s clever use of the 12 QPU-cores you found in every rasperryPi. See his work on “Python library for the 12 QPU-cores” (roughly speaking it is a QPU-Assembler, embedded in Python).

  3. crashoverride
    May 1st, 2017 at 10:58 | #4

    “Extraordinary claims require extraordinary evidence” — Carl Sagan

  4. galactic
    May 2nd, 2017 at 20:20 | #5

    Not sure how no source equates to a hoax…seems like conclusions based on limited investigation. As @Mic_s suggests his github accounts suggest that this has been a work in progress. Further investigation would show that he himself and his team are of reputable pedigree in terms of experience and education (top in Japan). They have a reputation to uphold.

    I agree accessible source code would solidify the claim, but I think we need to respect there may be reasons as to why such source at this time may not be released.

  1. No trackbacks yet.