Це відео не доступне.
Перепрошуємо.
Run Any Raspberry Pi AI Project FASTER! OAK-D Lite Camera, Getting Starting With Raspberry Pi
Вставка
- Опубліковано 5 сер 2024
- Ever needed a performance boost when running Machine Learnt AI Systems (like facial recognition) with a Raspberry Pi Single Board Computer? Or wanted some Depth Data from your Camera feed? Then the OAK-D Lite is for you!
Full Article (with Terminal Commands, 3D Printable Mounts and more AI Scripts!) - core-electronics.com.au/guide...
OpenCV AI Kit Depth = Oak-D
Related Information
Set Up a Raspberry Pi as a Desktop Computer - core-electronics.com.au/guide...
Face Recognition With Raspberry Pi - core-electronics.com.au/guide...
Object and Animal Recognition With Raspberry Pi - core-electronics.com.au/guide...
How to use Raspberry Pi Imager - core-electronics.com.au/guide...
Use Your Phone to Control Your Raspberry Pi - core-electronics.com.au/tutor...
This is a single USB-C powered module that sports 3 Cameras and a suite of circuitry. It is like a Google Coral but with auto-focusing cameras. It has an autofocus RGB high-resolution 4K central camera that can run at 60FPS and two 480P Binocular Vision (Stereo) Cameras that can run at 200 FPS. There is also internal circuitry to the Oak-D Lite that does the hard yards of machine-learned processing and provides the Raspberry Pi with the results. Thus this offloads the computational operations to the module and leaves your Raspberry Pi free and capable to perform other tasks. The brain of the OAK-D Lite is an Intel® Myriad™ X Visual Processing Unit (VPU).
OAK-D was the world’s first Spatial AI camera and OAK-D Lite is Luxonis's most recent offering. The two stereo cameras pointed forwards enable the OAK-D Lite to create depth maps and determine accurately the distance to identified objects. This is similar to human binocular vision, except much more accurate. They both can run multiple neural networks simultaneously for visual perception tasks like object detection, image classification, segmentation, pose estimation, text recognition, and more while performing depth estimation in real-time. Also, Happy Birthday to me! A number of these machine-learned systems I have explored before running directly on a Raspberry Pi 4 Model B, Face Recognition with the Raspberry Pi, Hand and Gesture Control with Raspberry PI, and Object Detection with the Raspberry Pi (to name a few). But none of these guides run at the high FPS (frames per second) or high resolution like the OAK-D Lite. This higher FPS speed unlocks new AI capabilities and ways to stack multiple AI systems in a single Python script that I have not been able to explore with previous hardware.
This guide will demonstrate exactly how this kind of AI processing can be harnessed by the everyday maker, whilst allowing them a peek in the programming back room to see and understand exactly what levers there are to adjust. By the end of this guide, you will have a Raspberry Pi single-board computer with a Spatial AI Camera capable of keeping up with the most current advances in Edge Machine Learnt technologies. Edge computing is the idea of pushing computing and data closer to where they are used, which means no cloud computing.
If you have any questions about this content or want to share a project you're working on head over to our maker forum, we are full time makers and here to help - coreelec.io/forum
Core Electronics is located in the heart of Newcastle, Australia. We're powered by makers, for makers. Drop by if you are looking for:
OAK-D Lite - core-electronics.com.au/catal...
OAK-D (Original) - core-electronics.com.au/catal...
Raspberry Pi 4 Model B (4GB) Ultimate Kit Bundle (AVALIABLE!) - core-electronics.com.au/raspb...
USB3.1 to USB-C Cord - core-electronics.com.au/catal...
Raspberry Pi Single Board Computers and Gear: core-electronics.com.au/raspb...
0:00 Intro
0:33 Oak-D Lite Overview
1:04 Depth Overview
1:35 OAK-D Lite on LEFT, Oak-D Original in MIDDLE, Raspberry Pi on RIGHT
2:07 What You Need
3:12 Assembly of Hardware
4:40 Software Set Up
5:55 Luxonis GUI Demo
7:59 GUI Demo Exploration
11:45 Available Open Source AI Systems
12:15 Remote Heartbeat Monitoring with AI
14:20 Face Blurring with AI
15:20 VR Projected CAD Control
16:12 Where to Now
16:49 Outro
Dude, owning an oak-d lite, this is a really good breakdown of this incredibly powerful combo, bravo 🤘
Really great overview! I appreciate all the relevant context throughout. ty.
Man this is great!
Just realized that here on the lab I work we have a robot with this camera attached. Gonna try to figure out how to make both work together
Very cool demo. Just getting started with mine and it was a useful overview.
Thank you.
Damn, that's a lot to learn, thank you. Haven't tried it but will do. Love from India
Actually, the Raspberry Pi 4 has all USB ports connected to the CPU using a single PCIe Gen 2 lane. That means a total of 4 Gbps shared across all USB ports. Even though it's correct to plug in OAK to the USB3 port, plugging other devices to other ports can still take away from the USB 3 port's bandwidth.
I found this while trying to design a multi-camera system.
Cheers for this new knowledge!
Great video. So many ideas swirling :)
Outstanding video, thank you so much for making this!
Cheers mate! 😊 Very kind.
Thank you for the videos you have been posting we are learning a lot from them.
Yeehaw 😊
No, kinect2 is a depth beast, the question which is better is not relevant, kinect for sure, question is if you need go outdoors and if you can power it.
OAK-D lite is quite poor in depth, but the IA thing is ok. kinect have not, so are oranges with apples comparison
Great stuff ! Really inspiring man
😊
Impressive!
Nice! Can the camera be used for pose estimation too? I saw your other video on pose and face detection. Would great to combine heart rate and pose detection for tracking and guiding workouts. Is the built in pose estimation here better than Open CV?
Immediate sub!🤯
I regret that I am only allowed to click the like button once! Muy bueno!
You absolute champ! ❤️ Thank you for these and all your previous kind words! Hope your New Year was great!
how can i use it to tracking an object and output to two X and Y servo motors to act like pan tilt , how many fps i can get , thanks
hi I would like to know how to make a video with a Rasp 3 + where I can also overlay information from different Arduino sensors so that everything is coordinated on a screen that is showing the HDMI output of the Raspeberry
Another question is posible split in multiple sectors the data in the video is for my project to a starship for my child thanks
Say i want to track animals such as deer, do i need an ai camera like this one or are there cheaper alternatives, like a regular pi camera to do the same thing just lower quality. Thanks 🙏
how can i find the best pre-trained models in open sourse
Which one is better using YOLOv8 and Raspberry Pi4 : Raspberry Pi HQ camera or that one in this video?
Can this camera detect the terrain on an image on TV screen?
Thanks for this video! Do you know the max depth distance by any chance? Is it more than a Kinect 2.0?
For OAK-D Lite the Maximal perceiving distance is ~19.1 meters. Source Material - docs.luxonis.com/projects/hardware/en/latest/pages/DM9095.html
Depth Range of Kinect v2 appears to be 4.5m. Hope that helps 😊
hello Brother, Could you help me in run my custom dataset which i already trained on yolov8 on this device?
I recently tried to follow the commands on the linked website and it says distribution is not supported when trying to download the install dependencies. Has anyone else ran into this issue and found a solution?
why is the audio choppy. artificially add in background noise so silence is not so jarring
Hi,
I’m using Raspbian GNU/Linux version 11 on my Raspberry Pi 4 und I just tried to set up
# OAK-D Lite like in the following article.
Integrated Computer Vision Package - OAK-D Lite With Raspberry Pi Set Up - Tutorial Australia
The first problem is that after the first command from the articele no Video Preview window pops up and further on no Graphical User Interface opens when I countinue the installation process.
The depthai-experiments/gen2-blur-faces and other experiments are working after the installation process but when I’ve a Phyton code like:
import depthai
from depthai_sdk import OakCamera
I also get the error "cannot import name ‘OakCamera’ from ‘depthai_sdk’ "
What can I do?
Hey, nice video. Just copped a luxonis camera myself and i am looking for a small tripod for it. Do you have the name of the one shown in the thumbnail?
Awesome hope you put it to good use! That tripod has been laying around on our desks for so long I'm not sure anyone remembers where it came from, if you google mini tripod or phone tripod I'm sure something similar will come up. Otherwise we carry two small tripods that would be suitable but not exactly the same:
core-electronics.com.au/portable-mini-tripod-360-rotation-ideal-for-raspberry-pi-hq-camera.html
core-electronics.com.au/the-pi-hut-flexible-tripod.html
wow this is so exciting! 🥳 I have a Raspberry Pi 4 order. Will the OAK-D Lite camera run on the Orange Pi 800 all-in-one-keyboard or the new Orange Pi 5? 😎 Thanks a lot.
I have no idea. But if you can run Raspberry Pi OS on those boards I reckon you would be able to. These systems have been designed to run on Raspberry Pi Hardware so don't hold me to it if things get weird 😛.
@@Core-Electronics Thank you.
I am not run OAK-D Lite on Raspberry Pi
So I occure some error, like Error 'module' object has no attribute 'freetype'. After change to stander cv2.putText the program can run.
But the weid thing is it always throw "Change exceeds pixel threshold, may need to switch to new data buffer", and I find out it can only be used for static, as soon as I start moving the bpm starts to Inaccurate
Heyya mate,
So are you running a Oak-D Lite with a Raspberry Pi or an Oak-D Lite with some other hardware? If you keep hitting the | Change exceeds pixel threshold, may need to switch to new data buffer | you can increase the Pixel Threshold Like I do in the video to the value of | 2 |. If your moving around a lot the BPM script will have lower accuracy.
Hi. How good is this camera to measure pothole's depth, installed anywhere in a car (let's suppose that the car does not ride very fast, +/- 20 km/h)
According to the data sheet for the Oak-D Lite, it's optimal range for accurate depth perception is 40cm - 6m, and it has a minimal Z distance of 20 to 35cm depending on usage. If you aim was to mount it underneath the car it seems like it wouldn't be possible with even the highest ride height of a production car being 25cm.
I'm going to write a program to sense who is looking at people's butts. Ima call it buttlooker. I called dibs.
Anyone experiment with a Coral USB Accelerator with this?
Hello How we can Create AI machine Vision using the Same Raspberry Pi and the TFT display with the same frame Rate
So long as you use the Oak-D Lite Module you'll be able to create a very formidable AI systems with a TFT Display. I would recommend using a TFT Display that uses HDMI.
@@Core-Electronics IT means with the use of Raspberry Pi Zero as a Computer and with the Same Computer vision Algorithm we can create a HIGH Frame Rate Tracking this will be a very good and nice high speed motion tracking very useful 👍
Proper exciting hey 😊
Can I control this OAK d camera using RaspContraller app?
Set up correctly most definitely! 😊
thank you your movies
please, control device for your voice because bass sound is high.
Cheers mate and I will look into it.
please make a video on pixycam with raspberrypi
PixyCam does look very rad! I've popped it into my to do list 😊
Hi, I get an error when I try to run the example test. I get ImportError: libcblas.so.3: cannot open shared object file: No such file or directory.
Is there anything you could suggest I do to fix this issue?
Hi! We're best able to provide technical support on our forums - there you can open a thread and include images and code snippets. head to forum.core-electronics.com.au/
@@Core-Electronics Hi thank you for the reply, I have solved the issue. I had to use the "fullest" version of Raspberry Pi OS. When using the less equipped versions of the Raspberry Pi OS it gives different errors throughout the installation.
Ah yes, good catch with that one. 👍
It's Data, not Data
There's no way your resting pulse is only 53 bps. A normal resting pulse is around 70 bps. And you are not really resting so the correct number could even be higher.
ooo really ?????
so this gentleman is lying?
All sensors that are worth while require calibration. So, sure a small mathematical tweak to zero it in would be expected. But this is powerful stuff. Great video.
Has anyone actually taken a look at this hardware or is the things he's saying pure marketing wank?
It is actually as shown. The demo was on real hardware, mine works just as shown.
It really is a great shame that your style of presenting irks me so much that I could only watch for a few minutes. I really don't know if all the dramatic pauses in speech are a bad habit or intentional but they really are annoying. Such a shame because the content that I did manage to sit through was very good.