Raspberry Pi AI Camera - Deep Dive

Поділитися
Вставка
  • Опубліковано 27 лис 2024

КОМЕНТАРІ • 27

  • @davidbayliss3789
    @davidbayliss3789 Місяць тому

    I thought that was a really comprehensive video, ta. Even an AI novice, so long as they've already earned some tech familiarity and reasonably able relevant intelligence, ought to be able to gain sufficient insight from watching this in order to understand what the new device does, in context, and devise ways of utilising it. Being able to communicate complex topics that well in a single session is a talent I perceive few to possess. i.e. making it appear accessibly "simple". It requires a teacher who not only has a pragmatically deep grasp of the subject they're explaining, but also a good grasp of students' levels of understanding and empathy for their perception and learning needs.

  • @DougToppin
    @DougToppin Місяць тому +2

    Very good overview of the camera and of how image processing works. tks for posting it

  • @jasonpocaro2730
    @jasonpocaro2730 15 днів тому +1

    Just got a pi 5 & m.2 AI top with 13 Tflop 😮 going to get a real good camera today and testing this weekend 😁🎉

  • @circuitfever
    @circuitfever Місяць тому

    Hi Kevin, this was a great presentation. Thank you for doing this!

  • @Unknown-si7go
    @Unknown-si7go Місяць тому

    Hello Sir, great explaination🤩
    I have a question, does the quality of the camera still same(fps) if I run a custom dataset with transfer learning tenchnique model like ssd mobile net tflite / yolov8. Or the best quality only for the models that has been provided?

  • @dermick
    @dermick Місяць тому

    Great video - really helped me a lot. My car's reverse camera does object movement detection - do you know if any models that do that with the RPi AI IMX500?

  • @bjy128
    @bjy128 Місяць тому

    Great video as always. Can it run Haar cascade models?

  • @TheOriginalLos
    @TheOriginalLos 3 дні тому

    You know ... this could put game cams out of business. Being a hunter i KNOW there is a way to get my Pi 4 and AI cam to be used outdoors (via solar panel power) to ONLY record / alert when a certain animal walks by. That would be a great business idea for people. Also to ONLY record "people" or "vehicles" for surveillance recording. I'd like to buy newer game cams these days with cellular network connection which ill have to pay for. But imagine having a PI with cellular, and AI cam + lense that will NOT record just everything nonstop but specific targets.

  • @robotboy3525
    @robotboy3525 26 днів тому +1

    Hi McAleer, would it be possible
    to get this Camera and the Hailo-AI
    acceleration module working together for more processing power ??

    • @kevinmcaleer28
      @kevinmcaleer28  26 днів тому

      Only one accelerator can process the model at a time, its either the camera's image signal process or the one on the Pi's AI Kit/HAT+ processing at a time.

    • @robotboy3525
      @robotboy3525 26 днів тому

      @kevinmcaleer28
      I see.. in this case would it be recommended
      to use the Hailo which has more processing power and a Pi regular camera or the camera AI has its advantages?? 🤔

  • @manfredbogner9799
    @manfredbogner9799 Місяць тому

    Sehr gut

  • @martinnash007
    @martinnash007 24 дні тому

    Interesting. What I don't understand is how a 24x24 pixel grid amounts to 180,00 possible features to test for? I would have expected around 424x424 to be in that ballpark! Am I missing something?

  • @lucasschneider9879
    @lucasschneider9879 29 днів тому

    how to use YOLO model on this camera?

  • @johnhamirkarim3696
    @johnhamirkarim3696 Місяць тому +1

    Can this accelerate face recognition?

    • @kevinmcaleer28
      @kevinmcaleer28  Місяць тому

      Sure - but you'll have to create a model to recognise a specific face. The model for face detection is very different from face recognition.

    • @johnhamirkarim3696
      @johnhamirkarim3696 Місяць тому

      @@kevinmcaleer28 hoping you'll make a guide for it🫰. Thank you for responding

  • @aikifab
    @aikifab Місяць тому

    So, no OCR features? I'm disappointed on that topic. Thank you very much for your very interesting video

    • @reinerheiner1148
      @reinerheiner1148 Місяць тому +2

      You should be able to load your own models, so training/loading an ocr model should not be a problem. OCR is one of the most basic machine learning tasks after all

    • @kevinmcaleer28
      @kevinmcaleer28  Місяць тому +1

      Which OCR model do you want to use? Pretty sure there are some standard models for that out there that can run on this without too much fuss

  • @TexasAi
    @TexasAi Місяць тому +3

    Hi! Texas Ai here! Would love to buy one. Any idea who has one in stock in the states?

    • @kevinmcaleer28
      @kevinmcaleer28  Місяць тому +1

      I think Adafruit will be carrying supplies of this as well as digikey and sparkfun

  • @Phonognomiks
    @Phonognomiks 21 день тому

    Anyone found a real use case for that? Like mapping on a screen on top of a human some character for example Santa or something?

  • @Eriiiiiiiick
    @Eriiiiiiiick Місяць тому +1

    FIRST

  • @JarppaGuru
    @JarppaGuru Місяць тому

    let say product AI name. no rush AI2 camera come

    • @kevinmcaleer28
      @kevinmcaleer28  Місяць тому

      No sure I understand the sentiment here.