We've attempted that method once before, his hand movements were too strong he nearly dislocated both his shoulders!

This is what I have for Raspberry Pi 5 Touch Screen, 3.5 inch Touchscreen with Fan, 320x480 Monitor TFT LCD Display

Thanks for the encouragement!

What Yolo version is the model? Not every model version supports the NCNN conversion with this method (for example Yolov10 doesn't). A helpful resource might also be the documentation page for this process: docs.ultralytics.com/integrations/ncnn/

Same, did you find anything?

What Yolo version is the model? Not every model version supports the NCNN conversion with this method (for example Yolov10 doesn't). A helpful resource might also be the documentation page for this process: docs.ultralytics.com/integrations/ncnn/

Unfortunately this is where it goes from somewhat-easy to quite involved as you have to get it in the specific format the HAT needs. Hailo has some documentation on how to do it over here, and you may need some decent hardware (like a gaming GPU, this can't be done on a Pi it needs a desktop machine): github.com/hailo-ai/hailo-rpi5-examples/blob/main/doc/retraining-example.md

This is by far the easiest too follow course/videos on the subject. And for some reason telling you so got deleted.

It's not currently on our radar, but I will add it to the list, no promises though!

No Wi-Fi needed just a LoRa module compatible with Meshtastic.

The Pico does not need one, we are bringing our own method of wirelessly communicating

We found that we needed more than 1GB of RAM, and it was already slow on a Pi 4. But if you have a Pi 3 already, give it a go! It might work!

That is some real out of the box thinking. I think there would be potential issues with pin and address conflicts using multiple HATs and I don't think it would work unfortunately 😔

We have a dedicated community forum topic for this video that might have some helpful information, if not feel free to post your issue! forum.core-electronics.com.au/t/how-to-make-a-raspberry-pi-nas-a-nas-berry-that-runs-open-media-vault/16580/8

I haven't used YOLO on a Jetson yet, but Ultralytics also has some really good documentation for it. I know they have really put in a lot of effort to get it going on a Pi and because of that its easy to set up and load custom models onto. I'm not 100% sure but I assume it would be reasonably easy on a Jetson as well. The Pi 5 will also likely be a cheaper option, both in how much the hardware costs, but also power consumption (a 4GB Pi 5 will work fine here). But the Jetson will obviously be way more powerful and give you greater FPS. I think it comes down to the question of "will the Pi 5 run the model well enough for your needs?". If it does, go with it. If it doesn't, then maybe give the Jetson a go. You can get some really fast FPS on the Pi 5 if you lower the resolution, but this limits the distance you can detect objects at, and sounds like you might need some range on your project. However, you should be able to run your model at 4-5 FPS at full resolution if that works with your project. Hope this helps!

​@@Core-Electronics Thanks for your response, it was really helpful. I'm considering using just one video segment and reducing its resolution to improve the model's performance on the Raspberry Pi. There is a specific region I want to monitor, and the other video segments are not necessary for this purpose. Additionally, a lightweight model should be sufficient to detect what's a bot and what's not, ensuring an efficient and effective solution. I really appreciate your help!

No worries! Never thought about using only 1 section in a video, if we could redo this guide again we might of included it!

In the written guide linked below (also here):core-electronics.com.au/guides/raspberry-pi/getting-started-with-yolo-object-and-animal-recognition-on-the-raspberry-pi/

@ me aparece que la página o el foro no existe

Im having problem with the cv2. Imshow and qt5

We're glad you liked it!

Hi, that's an excellent question and I had to look into it myself as this video is a bit old. From my understanding, this product has various regional versions, each with slightly different product names. The G-H version we are using is the global version, and there isn't actually an X version, so my logical assumption is that Waveshare produces all variants with X on the silkscreen to denote a wild card meaning they don't have to manufacture different products for each region and only need to tweak firmware.

Follow up - the regional designation G-H is visible on the center chip, the X is on the PCB silkscreen to limit unique parts in production.

@ thx for clarifying

The model by default should be based on Yolov8, but the size of the model may depend on the HAT you are using, it will either be Yolov8m or Yolov8s!

We have been having some rare cases of it not working, if you would be willing to help us out could you let us know what browser you are using? Also this link may work: core-electronics.com.au/guides/raspberry-pi/raspberry-pi-kiosk-mode-setup/

@@Core-Electronics I've tried three times to reply, but for some reason my reply is being deleted. I am using chrome (latest update). My access is denied, error 1005

Sorry to hear that mate, thank you for the error code, ill pass it along and we will try and fix it.

At the bottom of the page is there a Ray ID? We can specifically check out your request when you try and access the page