Raspberry Pi Zero + camera vs. OpenMV

Project Discussion
#1

Disclaimer: I am OpenMV original kickstarter supporter

I am evaluating new project feasibility in image processing. The project requires fast deployment with portable product and user convenience. The list of requirements is longer, but let us focus on these.

In 2018, there are two options:
a) Raspberry Pi Zero W available with camera support
b) OpenMV7

Raspberry PI zero offers following advantages:

  • lower price
  • comparable power consumption on power saving mode
  • Bluetooth + wifi available (higher power cost)
  • linux with community support on various software
  • small form factor,
  • camera on ribbon, various cameras can be exchanged

I see the following advantages of M7:

  • realtime operation
  • easily programmable
  • low startup times
  • all in one package, except wifi
    +++ open micropython

Question:
What are the disadvantages of Raspberry PI and OpenMV M7 as compared to each other for a portable machine vision product?

Mostly interested in minimizing time spent developing the hardware and software solution for still image capturing and processing at a low frequency, i.e. couple of frames per second, and displaying the resulting graph on an arbitrary color display.

My experience with OpenCV is limited, and I have not tried Raspberry PI extensively, nor the new model M7 of OpenMV, so all advice and experience is welcome.

#2

Hi, I’ll write an in-depth reply to this later tonight (12 hours from now).

#3

Hi, here’s a detailed answer to your question:

Question:
What are the disadvantages of Raspberry PI and OpenMV M7 as compared to each other for a portable machine vision product?

Answer:
The OpenMV Cam has an edge over the Raspberry Pi Zero when you consider the value of your time in developing an application. In particular, if you are trying to implement a feature for an application which the OpenMV Cam’s firmware currently supports. The OpenMV Cam offers a much quicker solution to your goal than any other system. In fact, this was the whole point of the OpenMV Cam. Ibrahim and I brought the camera into existence for the purpose of enabling folks to more readily produce applications using computer/machine vision without having to have extensive Linux experience in-order to use OpenCV. The OpenMV Cam and OpenMV IDE both support this goal.

That said, at the moment, the OpenMV Cam is not as powerful as the Raspberry Pi Zero. Our, next version will close/exceed this gap and add global shutter support which the Pi does not have and which is necessary for serious machine vision applications (and after that we’ll like to do a linux based camera which is backwards compatible). In the mean time, if you are trying to decide between one and the other… it really depends on the application.

…

If you plan to implement high speed color tracking we can offer a much higher and lower latency frame rate output than the Pi. However, we have pretty poor WiFi/Bluetooth support. That said, you can use the OpenMV Cam with an ESP32/Particle easily and just make the camera an attached serial device. For your case, you mention a low frame rate application. What in particular are you planning to do?

#4

Hi,
I am also an OpenMV original kickstarter supporter.

I’m trying to build a system with an openMV + keyboard + LCD + scale, that:

  • recognise food packages (milk bottle, meat package, grosseries, etc)
  • recognise bar codes,
  • add that to the scale (weight) info and
  • send info of in/out to a database,
    this will allow me to know what I have in the fridge/storage.

Is this possible with M7 ?
What do I need to do to allow it to recognise all those food Item?

I’m starting to learn and use Python, ML, OpenCV, etc.

Best regards

#5

Hi, you’re probably going to want to use a Pi if you plan to do all that. The OpenMV Cam is really best suited for small robotic applications where the focus is to build something that works as quickly as possible. From what you sound like you want you’re going to want the full desktop computer / pi experience.