Unlock AI Power | HAILO AI Accelerator on Raspberry Pi - Beginner to Advanced Guide

[Music] [Music] hey guys welcome back again to this video of Elon explanations so today we'll be unboxing this AI kit first we'll take the look at the AI kit so which comes pre-installed with the M2 hat plus key so as you can see there I've also installed that heat SN and that's the PCI connector and you can just connect it there for the Raspberry Pi which is fully compatible and that's the heat SN which I have installed again it doesn't come pre-installed with the AI kit that's the box and uh we'll go through some get started links and features and warnings of this AI kit it's clearly written there if you want you can please check it out that's for the box and then we coming to the M2 hat plus key this is the overview of Halo AI module and uh we'll also see whether it is completely compatible with this Raspberry Pi 5 and that's the lock pin there so you can just lock it there by pressing that black color key there so you can just put that PCI connector and then lock lock it there it also comes with a four column stand so that's the first one one and the second one 1 2 3 and four that's the heat sink and a cooler cooling system for the Raspberry Pi 5 it doesn't come with the Raspberry Pi 5 you you have to buy it again and those are the screws to fit the AI module so you'll actually get 12 of them I've just shown four of them here it's better to use the longer ones on the top so because you have to again connect the GPS that's the four shorter one at the bottom that's with it you can keep it on the top and then connect it there and then set up the screws and then connect it to the power supply and then turn it on that's with the installation process of Hardware so don't forget to use 27 wat official power supply for this a module because it draws some external energy and the worst design of this is that gpios a connector thing it doesn't completely connect you have to remove the Hat when you are trying to use gpios because those gpios doesn't come out of that AI uh kit M2 hat that's the worst design and you cannot even leave it in the midair it doesn't completely connect those are the holes which I'm talking about the GPS doesn't come out of there and they are light fragile also they'll even bend easily if you try to bend them so this with the hardware setup and that's with the GPI opens those are light fragile and then next we'll go to this set software installation and a theoretical explanation process before we completely set up this Halo a mode before moving to download the software installation for the Halo module in Ras b os we'll get to know some of its Futures in its official data sheet which is more important because you have to get to know about the processor in detail before you customize it use it for your custom projects okay this model is ATL it's an entry level AI accelerator this is more different than a normal processor it will just enhance your operations instead of completing the operations completing the operations will be in the hand of your host processor it'll just give uh give you a helping hand for your host processor let's look at some of its key features and benefits it is capable of doing 30 13 tops where the Google Coral AI accelerator is just doing uh six stops out there and absolutely it'll consume low power when compared to other processes out there because you know for the Raspberry p you can't expect it to consume more power than required but when you are doing some complex task it will consume a lot but in general it doesn't when you compare with other process out there and these are some external features and benefits so yeah easy Hardware integration is one of them and you can operate in dry and medium temperature atmosphere also that's not a problem key benefits of this are uh detection of fast moving objects yes it can do that and high accuracy and ability to detect small objects I don't think so high accuracy is okay but detection of small objects nope it did not work for me so I don't think it is right now capable of doing this but it might be in future they have to upgrade some of the packages which comes with it but yeah it can do it in future and low latency and real time alerts yeah low latency can be achieved when you use an external SSD or you know nvn I think you can't use nvme because you're already using your hat for your uh haloi but you can use an SSD if you want efficiency it's fine but cost saving no I don't think so because it is more than my Raspberry Pi module it it costs around 7K where my Raspberry Pi 5 that with 4GB module costs around 5,600 in Indian rupees so I don't think it is cost saving when you compare it with raspberry budget thing but when you compare it with other processors out there it is somewhat cost saving this comes with uh three model things which one of them is chip on board which will be using Halo ADL and uh these two are different so this chip is here right so this comes pre-installed so which is this one Halo ATL M2 key A+ C so this is a chip on board design and uh this will be installed for our uh Raspberry Pi it will come pre-installed don't worry you will also get your thermal paste beneath it so it is thermal efficient ient also this is its uh build and runtime environment this is the host processor our broadcom processor you can clearly look at this if you download the data sheet I'll give you the link in the description have a look at this I'll explain you in the future what is this St AS application examples and yeah it is also capable of neural network models and application support this is important how it uses with our uh system and uh how it does connect with our system so this is the host processor your broadcom processor and there is the Halo atel which is being connected bya PCI connector this is this run environment right these this uh this section this part of Port processor is for user application some memory block allocated for that and another memory block which is connected via PCI or ethernet will be Halo atel device which comes with Halo ATL firmware it will be like this when you take a one block representation yeah this is with it and uh we'll go to the GitHub repo we'll see some of the examples which this is capable of doing and it'll literally blow your mind because you won't think that Raspberry Pi is capable of doing these things this is the GitHub repo which they have created recently so if you come here and then go to the read me file over here I hope this is visible right and they've named this Raspberry Pi 5 example [Music] Labs if you come here this is how it looks uh this is without the heat sink over here I've just applied a heat sink [Music] here yeah if you want a detailed installation guide I can give you that I mean this is is slightly you know Compact and they have reduced some of the instructions you need to follow in between but you can also use this if you want more in detail just let me know in the comment section below I'll create a GitHub repo in just one or two days and then I'll provide you the link there or else you can just find all our GitHub repos and all our projects in GitHub if you just four car repositories I'll also provide you the link there this is one of the example so this is the detection example which detects humans objects or you know the one which it knows so for me it did not detect airpods and uh it did not detect my screwdriver and if I pointed my camera to the light it gave it like a airplane airplane thing so I don't think it is fully compar with all realtime objects but they have programmed some of the objects there like baseball bat baseball glove and a Truck a bus a car a person for sure even if it you know look at a dress of yours it'll identify it as a person around 30 to 40% this is a pose example po estimation example even this works fine when you you know keep a distance with your camera I'll also show you the quick brief of Raspberry Pi camera module 3 and this is instant segmentation uh example this typically Works in apple you know portrait mode photography you know it just keeps the person or the object detected in the more uh Clarity one and the background blurred so we can get more focus on the person or the object we are looking at and you can just Implement them by typing the code in your raspberry pii terminal before creating that you know the code in your terminal you have to create a virtual environment before that you have to clone this Repository and before that you have to upgrade some packages and you have to configure your Halo module which will'll be doing right now in here if you go to this Dock and then install Raspberry Pi 5 thing this is the detailed installation guide for your Rasberry Pi 5 so first they have installed using an maor just have installed a Raspberry Pi OS you know keeping in mind that you did not but if you did not install Rasberry Pi OS you can just view our first video in this playlist on how to install raspberry pios I have explained in detail for all operating systems if you have any doubt regarding that also please comment in the comment section below and also you can connect to us in the Discord we can we can discuss in detail over there to install Halo software you have to type this command sud sudo app install Halo all we'll be going this in detail you don't need to worry about all this and also after installing everything to make sure that uh your Halo is completely connected you have to run two to three commands the first first one is this one yeah before that you can yeah before that you can run this one to identify that your Halo HL is configured properly if everything is okay the output will be something like this if you get an error please comment that we'll look at it and this is for inspecting the tools which comes with Halo AI module they have also given you some troubleshooting tips and uh they have also given you some known issues here you can go through them if you encounter any of this in your raspberry pii I don't think you will encounter but if you do encounter then please look at this in detail the first thing before getting into the software installation setup is that the one you have to use the original power supply and the second one make sure that your SD card is above 64 gigs because 32 gigs I don't think it will be sufficient for for the projects also for installation it might be sufficient but for you know for further uh packages and its installation I don't think 32 gigs will work out please make sure that you are at least using 64 gigs and your RAM must be at least of 4 GB if you're using earlier versions of raspberry pi5 I mean 4 three or such things I don't think you will get a PCI connector if if you do get the PCI connector that's fine yeah and also sufficient internet speed with these all let's get started with the software installation in Raspberry Pi OS now we'll see how to use Halo AI in Raspberry Pi OS I've have tested all the browsers like Chrome Safari Firefox and brave out of all them to start Raspberry Pi connect Brave the lightest one you can use that and it's also free for download I'll provide the link in the description so what is this RPI connect raspberry connect if you are having trouble finding the IP address of your Raspberry Pi or your if you are having trouble using your VNC viewer then you can use this official Raspberry Pi connect from raspberry before doing that you just have to connect to your raspberry via SSH if you don't have an external monitor if you have an external monitor please skip this part after connecting to your raspberry via SSH you have to pass on this command and then your screen sharing from your Raspberry Pi application will start and then come here to the brave click on this you can connect via remote shell or screen share remote shell is nothing but you'll get another terminal window over there but we'll use screen sharing right now because we need the interface of desktop to install packages and run them if you want to run your camera package and your detection modules you need to have your desktop turned on so it will take some time like around a few minutes later 30 seconds or 40 seconds until then wait I'll tell you how to use this just search Raspberry connect and then sign in with your account raspberry account yeah I think we are ready to use it now this is our raspberry window so without VNC view are we are able to view the desktop also this is an useful application I have found out this recently yeah yeah apart of all that let's start our process you all connected via screen sharing using Raspberry Pi connect or VNC VI or you might have your external desktop but if you are connecting via screen sharing open your terminal and then type these commands first you have to start with sud L AP update by the way I'm using remote connect so that I can increase my font like this and it will be clear for you first type this the system will automatically update and then the second command is Pudo appt upgrade and uh Pudo app not upgrade it should be sud sudo appt full upgrade okay this is the second command it will upgrade all your packages all your external packages also and the next main thing here go to sud sudo raspy config in here you have to go to Advanced options and here select PCI speed and then set it to gen 3 make sure you do this because we need maximum capacity from PCI Port okay I'm not doing that again come back after selecting that press back click the right arrow to select the down options the select and finish click the right arrow twice to select finish and then press enter and then type Pudo reboot it will reboot your system just wait after that now here comes the special command Pudo app install Halo all this is the command to install all of the Halo AI packages type this wait for some time it might take time if you are with slower Wi-Fi network it will all included with this package whatever you require for your Halo projects and then reboot your system again to apply these changes verify your installation by pressing this the output should look something like this it should give your device architecture serial number part number and your product name okay and we have to also include some of the Halo AI tools okay I'm sorry all are included in that Halo all installation command so don't worry everything is installed we have to check them verify them to verify them to verify whether Halo tools are installed or not just press this command GST inspect 1.0 Halo tools and then click enter your output should be something like this if you are facing any kind of Errors when you press these two commands hello tools and uh your identification of your Halo module please comment down in the comment section or join our Discord channel for further connection just type Q to quit that and then you'll be back with a blank after downloading Halo tools and inspecting them also we'll next check whether the Halo RT interfere G streamer element is installed or not so for that click this command and press enter you should again see an output something like this your version your license and name description F right yeah after that everything is installed you're good to go just one command which will download all the required Halo packages and then come back here to your browser go to GitHub click on dock and one more thing here is your installation thing if you are facing any kind of issues like PCI troubleshooting or driver issue or there are some known issues over here like PCI page size issue or cannot allocate memory in static TLS block issue this is a very important one don't neglect it don't move ahead if you are facing something like this and you are forcefully trying out some commands again and again and again don't do it come here explore them and solve your issues if you find any some further difficulty please feel feel free to contact us go back to dog here and uh go to basic pipeline. MD now to check out some of the crazy Halo modules which are pre-installed just copy this you have to clone the repository first come back to our shell type that command [Music] here I already cloned it so it should look something like this uh Halo RPI examples after cloning it will be something like this when you type LS you'll get all your directories in your user one and then type CD H tab and then click enter type LS to list out the items in this repository you'll get all these basic pipelines readme doc if you go to Doc again press LS here you'll find basic pipes. MD installation thing and some of the majes these are text files basically you can also inspect them with cat command go back from here first you have to set up your virtual environment to test out some modules with this so for that what you can do is type Source SE and then tab it aut automatically type it for you press enter this script will be executed and you'll be in Virtual environment Halo RPI 5 examples if you wonder what's there in that script and what what it is doing in the background we can inspect it click Nano setup and then you you just click SC and then tab and then open it so they have coded something for us they have basically written a script to to start a virtual environment first they have written a function to check if the script is being sourced and then only proceed if the script is being sourced get some of the versions of Tapas and get the directory of current script first it'll check if the virtual environment exists or not if it does not exist then it will create a virtual environment see virtual en a virtual environment does not exist creating one and it and then it will create a virtual environment for you after creating a virtual environment it'll give you some message something like this let's go back yeah activating it will basically go in a loop for after setting a virtual environment it'll again check for a virtual environment whether it exists or not if it is exist so it'll give you a message virtual environment exist activating it it'll activate the existing environment and then you'll be here now what you'll do is we have to type some more commands to install more resources type this command pip install minus r requirements.txt I'm not typing it because I already have it it if it does not give you you know installation thing because pip is an externally managed thing you can also use it with Pudo apt or your terminal will give you some suggestions on how to install that particular txt file and then you have to click this one from your GitHub just copy that command and then paste it over here download resources. and then it will download some of them and after that you are really good to go first we start with the detection example so if you want to run this you cannot run this via remote shell or your terminal you are uh you are using your uh Raspberry Pi OS via screen sharing or your VNC or your desktop right so on that type this command not this one yeah this one python basic pipelines detection. py it'll take input from a video file and it'll run the frames and then capture the details in that video file you can give a video file or you can just activate a cam also your camera also to activate the camera just delete this resources detection and then type RPI to start your camera before starting your camera make sure all the packages related to the camera are installed you can just type Pudo app full upgrade again to check whether all are upgraded and up to date it'll update everything for you including camera package also no need to worry need not search for uh another camera packages in the internet it will download everything for you don't worry this is one of the detection examples there's a detection one a model detection one you can also use a POS estimation example the second one in that guide which is this one running this command will give you a pose estimation example again it'll take a detection video a sample video if you want to start your camera or if you want to input any other video for suppose let's say if you want to start your camera just type RPI delete that MP4 thing just type RPI after input or if you want to you know load up your own video file then go to resources and then from there you can just copy your video file of yours and then paste it in the resources directory you can paste the name of that particular video file and then type it here you have to basically copy that to this Resources directory right and then after that you can paste the name right here okay and there's one more thing which is instance segmentation example this is also one of the example which they have created for a demonstration purpose these are the three basic examples which you can run I'll also show you the outputs if possible with pictures this is it this is with the basic installation of Halo AI software Hardware installation and the ingredients which we require to build some projects with this if you want more projects related to this haloi module just comment down in the comment section I'll provide you all the resources that you need to buy or if if you are doing some kind of simulation also I'll provide you the links or websites to that you can just read the data sheets from there you can do your experiments according to that that's it for this video and then I'll see you in the next one until then signing off wait wait wait wait don't forget to subscribe our channel for this exclusive and premium content just for you and it's the latest one in the market so it's just released before for 1 month or 20 days I don't know but it's the latest one so please make sure you subscribe and hit the Bell icon there until then signing off [Music]

Channel: HelEx

Views: 1,607

Keywords: #HAILOAIAccelerator, #RaspberryPi, #AI, #ArtificialIntelligence, #MachineLearning, #DeepLearning, #TechTutorial, #AIProjects, #AIOnRaspberryPi, #TechUnboxing, #RealTimeAI, #AISetup, #TechReview, #DIYTech

Id: bDkYTdHa26U

Length: 33min 25sec (2005 seconds)

Published: Tue Jul 02 2024