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.
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?
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?
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.
@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?? 🤔
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.
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?
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 hello, so would it be possible to have a face detection model then feed it to face verification model? Any models you can think of for this use case?
Very good overview of the camera and of how image processing works. tks for posting it
Thanks
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.
Hi Kevin, this was a great presentation. Thank you for doing this!
Just got a pi 5 & m.2 AI top with 13 Tflop 😮 going to get a real good camera today and testing this weekend 😁🎉
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?
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?
Hi McAleer, would it be possible
to get this Camera and the Hailo-AI
acceleration module working together for more processing power ??
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.
@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?? 🤔
Great video as always. Can it run Haar cascade models?
Sehr gut
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.
how to use YOLO model on this camera?
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?
So, no OCR features? I'm disappointed on that topic. Thank you very much for your very interesting video
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
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
Hi! Texas Ai here! Would love to buy one. Any idea who has one in stock in the states?
I think Adafruit will be carrying supplies of this as well as digikey and sparkfun
Can this accelerate face recognition?
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.
@@kevinmcaleer28 hoping you'll make a guide for it🫰. Thank you for responding
@@kevinmcaleer28 hello, so would it be possible to have a face detection model then feed it to face verification model? Any models you can think of for this use case?
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?
let say product AI name. no rush AI2 camera come
No sure I understand the sentiment here.
FIRST