The New OpenMV Cam H7

Поделиться
HTML-код
  • Опубликовано: 5 май 2018
  • This video shows off the new features of the next generation OpenMV Cam H7. It's powered by the new STM32H7 microprocessor running at 400 MHz with a Coremark score of 2020 - comparable to the 1 GHz Raspberry PI zero 2060. With the OpenMV Cam H7 hobbyist will finally have access to affordable global shutter cameras for precision motion capture computer vision along with easy to use thermal computer vision.
  • НаукаНаука

Комментарии • 31

  • @baetenj
    @baetenj 6 лет назад

    This is really incredible. Can't wait until it is available.

  • @outofthebots3122
    @outofthebots3122 6 лет назад +1

    Awesome work guys.

  • @aleksjal3593
    @aleksjal3593 Год назад

    Best vision sensor from top to bottom, thanks guys for such a great work!

    • @OpenMVLLC
      @OpenMVLLC  Год назад +1

      Thanks! Make sure you are on our news letter! We will announce the next gen platform on Monday!

  • @Tutoelectro1
    @Tutoelectro1 3 года назад

    Hi, I just discovered your camera. It is great! I will try to save some money so I can do some tutorials on my channel. I might even be able to use it for work. We are using cameras to make the PCR detectors that are so common now with Corona. Have a nice christmas!

  • @viney50567
    @viney50567 6 лет назад

    In your email you say the next firmware release will be able to run CNNs (which is awesome!). Will it be able to run RCNNs as well?

    • @OpenMVLLC
      @OpenMVLLC  6 лет назад +1

      The planned initial support is just for history less feed forward networks. However, it should be possible to allocate a buffer for history and use that again. Not sure if we will have support for this at first.

  • @ondrejzvozil2822
    @ondrejzvozil2822 4 года назад

    Hi, first of all, I am very glad I found this amazing project. I am thinking of self-made trail camera, but starting from a scratch is nonsense for me. Before I saw this project, I had also an idea that it might be enough to interface Omnivision camera with STM32H7. But, according to basic bit-rate calculations, I came to a conclusion that external RAM is needed (and of course I do not have knowledge of designing such devices). Is your device capable of storing 60 seconds clips on SD card in 1080p @ 30 fps? Thank you!

    • @OpenMVLLC
      @OpenMVLLC  4 года назад

      Our new OpenMV Cam H7 Plus can take 1080 images. It cannot achieve 30 fps however. The STM32H7 processor isn't really designed to do video. The camera can send JPEG images to the H7 and that can be stored to the SD card as an MJPEG file. However, you'd need to modify our firmware to have a fifo buffer to get smooth video. Taking snapshots is fine however. Note that our product is designed for machine vision. We have no video fifo buffer onboard. This is on purpose such that when you process and image it's the latest image and not delayed by several frames leading to response lag. If you are trying to build a video camera there are other processors more suited for that. Our system is mainly about processing on very low-power, cheap, and easy to buy hardware. The STM32H7 is sold on digikey without any NDAs needing to be signed to get it to do anything.
      Anyway, if you are okay with taking 5 MP snapshots the OpenMV Cam H7 plus will do what you want. It can also do MJPEG video. However the firmware is not setup with a video fifo so if you try to record an MJPEG video it will be jerky. If you modify the firmware to buffer frames when writing to the SD card you can then achieve smooth video.

    • @ondrejzvozil2822
      @ondrejzvozil2822 4 года назад

      @@OpenMVLLC Thank you for your extensive answer. I have expected that STM32H processor is not designed for processing the video of such resolution and frame rate. Could you please recommend me other "free to use" cheap processor that could handle this task? OR maybe similar ready device as yours? I found nothing. I recently inspected my low-cost trail camera and there is some low-cost "still camera processor" (NT96223), but of course it is not available for end-user and there is lack of documentation... And of course, it needs to run on batteries, for a long time, and it needs to "wake up" and start recording in a fraction of a second, so raspberry is out of the game...

    • @OpenMVLLC
      @OpenMVLLC  4 года назад +1

      @@ondrejzvozil2822 This is why I'm working on OpenMV, to make this possible. Right now if you want high end features you have to be building millions of units to get access to high end chips. That's just what it is right now. The point of OpenMV is to get to a future where camera data processing is not locked I to high end chips only.

    • @ondrejzvozil2822
      @ondrejzvozil2822 4 года назад

      @@OpenMVLLC Thank you for your work, I will stay tuned :)

  • @mariocenteno1
    @mariocenteno1 5 лет назад +1

    Would be cool to see this running on I.MX RT

    • @OpenMVLLC
      @OpenMVLLC  5 лет назад

      One of their application engineers already did a large porting job. I expect they will release something after the H7 is out.

    • @andrsam3682
      @andrsam3682 4 года назад

      Today I was at NXP's seminaire where they told that they are preparing release of new i.mxRT1170 MCU with 1GHz core clock. That will be really interesting...

  • @dcbonilla10
    @dcbonilla10 3 года назад

    It is possible to program it with C/C++? Or just Python?

    • @OpenMVLLC
      @OpenMVLLC  3 года назад

      Yes, github.com/openmv/openmv/tree/master/src

  • @AnthonyWebb7
    @AnthonyWebb7 6 лет назад

    Dude, tremendous! You selling these?

    • @OpenMVLLC
      @OpenMVLLC  6 лет назад

      We'll be doing the KickStarter for these later this year.

    • @mrtmrf5007
      @mrtmrf5007 6 лет назад

      OpenMV, LLC searched to get these...couldnt find them...lol...let us know when in sell

    • @OpenMVLLC
      @OpenMVLLC  6 лет назад +1

      As stated in the video, the plan is to launch a KickStarter later this year.

  • @Elamin008
    @Elamin008 6 лет назад +1

    any AI implementation in future ?

    • @OpenMVLLC
      @OpenMVLLC  6 лет назад +1

      Yes, please see our openmv.io blog post.