Ep06: Machine Learning Environment with Quarky Intellio | Using PictoBlox

Duration: 10:40 1438 transcript words EN summary

Video summary

In this episode of the Quarky Intellio series, the host introduces Door Bot, an AI application that uses machine learning to recognize faces and unlock doors, setting the stage for a deeper dive into creating custom projects. The main focus of the video is demonstrating how to build a gesture-controlled presentation system using PictoBlox's machine learning environment. By training a hand pose classifier to distinguish between left and right gestures, the project allows users to navigate through PowerPoint slides simply by waving their hands, effectively turning physical movements into digital commands for advancing or going back in a slideshow. The tutorial walks viewers through the essential steps of creating and training this model, starting with defining two distinct classes for left and right hand poses. The process involves recording various examples of each gesture using a webcam to ensure the AI can accurately identify them, followed by a brief training phase that prepares the model for real-time detection. Once trained, the model is exported to the Intellio board, which connects via Bluetooth to a laptop running on a 2.4 GHz Wi-Fi network. The video details the connection process, including verifying the serial number and observing the LED indicators that confirm successful pairing between the software and the hardware. To bring the project to life, the host integrates the trained model with PictoBlox coding blocks to manage slide transitions. A variable is created to track the current slide number, allowing the code to increment or decrement the backdrop based on detected gestures while preventing navigation beyond the first or last slide. The initial challenge of having the camera feed obscure the presentation slides is solved by adjusting the transparency settings within the machine learning environment, ensuring the slides remain visible while the AI continues to analyze hand movements in the background. The video concludes by showcasing the fully functional system where users can seamlessly control their presentations with simple hand signals, highlighting the versatility of machine learning for applications like smart attendance systems and currency detection. The host encourages viewers to experiment with other creative projects such as object detection or face recognition using Intellio, inviting them to share their own project ideas in the comments. Looking ahead, the next episode promises to explore four different methods for connecting with Intellio, helping users choose the best approach for their specific needs.

Read the full video transcript

Hey, hey, have you met our new friend door bot? All right, [music] he's talking about the door with just recognized my face using ML and unlock the door. So, hey, hey everyone. This is Abi and welcome back to our quirky Intel series. In today's video, we'll explore how Intel can see, recognize, and detect patterns using the ML environment in picto [music] blocks. And it can also take actual decisions just like real world AI systems. Just like this smart attendance system that recognizes each individual and records the attendance in Google sheet automatically [music] or even gesture detection to control your presentations or currency detection. The possibility are endless. Now all these projects may look different but [music] they all are built using machine learning which means learn from the data then [music] decide and take action. So now let's build something super cool. I'll show you how you can control your presentation using hand gestures. First, let's open Pictoblocks and choose machine learning environment over here. [music] We'll click on new project and let's give a project name gesture detection and I will choose a hand pose classifier [music] over here as we will be using gesture to control the presentation. Then we'll click on create project. Once this is created, we will be creating two classes. One [music] for left gesture and one [music] for the right gesture. With left [music] your presentation will go in the previous direction and for right it will go in [music] the next direction. Now I'll open webcam and let's train it for the left gesture. [music] First I will click on hold to record. Give all the possible left gestures. I'll try with this hand also. Okay, I think that much is enough. Similarly, now I'll do it for the right side. Okay, I think these images are enough. Now, next step will be we can train the model. It will take few seconds to get trained. [music] Now we can do the [music] testing. Let's click on webcam. Now let's do the left side first. Yes, it detects. Similarly, if I do the right. Yes, it detects [music] it properly. Perfect. Now I will export this model in block. And now we'll start the coding. So as a [music] first step, we'll connect Intel with picto blocks. So I'll switch on Intelio. Then click on board and select Intelio as a board and choose the first Wi-Fi router option. Now enter the same Wi-Fi name and password to which [music] your laptop is connected currently and make sure your Wi-Fi is running on 2.4 GHz [music] band as Intel supports only 2.4 GHz network. Check if your systems Bluetooth is turned on or not. [music] If it is turned on, great. Now is the next step. Click on next. Do select your Intelio serial number which is next to the power switch over here. Mine is AD2A. Intell's RGB LED will turn green as you can see on Bluetooth [music] connecting successfully. Now it's RGB LED will blink yellow while connecting. Once the connection is successful, you will see the RGB LED glow yellow [music] continuously. Then click on go to editor. Click on the camera icon over here. And you can see Intellio's camera view. It's an indication that we have successfully [music] connected. Now for coding, we will first get an event when green flag [music] clicked. Now first thing I want to open the recognition window. Okay. [music] So over here I'm going to put recognition window. Now uh how am I going to change the PPT? Okay. So I'm I'm going to upload my PPT in the backdrop actually. So I'm going to hide the Toby and I'll upload the backdrops. >> [music] >> I'll arrange them in a proper sequence. So first slide I want it to be the welcome. Last slide should be the thank you. Then each riddle one by one. So I can name the first slide over here as backdrop [music] one. Similarly you name the others as well. Perfect. Now I'll go back to blocks. Now over here first my backdrop should be this welcome screen. So what I'll do is I'll create one variable and give it a name as backdrop and I'm going to set this variable to backdrop one. So initially my backdrop will be one. [music] I'm going to use this variable and map it to the backdrops over here. So [music] I'm going to put a forever loop and in that I will [music] analyze whatever we are showing in the camera. So analyze the image from the web camera [music] and in this there are certain conditions which we're going to put right. So on the left gesture which PPT should change [music] and on the right gesture how it should change to the next one. So if condition like this now my first condition should be first of all if any pose is detected or not. So if the pose is detected then [music] I will check another condition if it's a left pose or right pose. Right? So one more if condition go to control and I'm going to put if and else in this. So if if [music] the pose is detected, if it's right, then what do I want to do? I want to change the backdrop [music] by one. So first of all, I'm going to change that to the variable. [music] Change the backdrop by one. Why we using variable? So that it can increase the number. So when I say change, it will increase the number by [music] one. And same thing, I'm going to put it in the backdrop over here. switch backdrop to and I'll place this variable. Perfect. And we'll [music] wait for 2 seconds before it detects the next pose. [music] Okay, it has to wait for 2 seconds. Now, exact same thing. I will duplicate and put it in the else condition. [music] The only difference is instead of right, it will be left now. And over [music] here first I will put a another condition that if the backdrop is greater than one then what it will do? It will go to minus one. Okay otherwise it should not go to the previous one. Right? [music] So if we'll put an operator over here let's say operator we'll put a greater than over here that [music] if the variable backdrop is greater than one which means it is going to the next backdrop only then it will go to the previous one right so in that case I will change the backdrop by [music] minus1 like this and same thing I'm going to put it over here. Switch backdrop variable. Wait for 2 seconds. [music] And as always, I want as always I want the backdrop to be whatever backdrop we have we are showing on the [music] screen. According to the gesture, it will keep the current backdrop. So I feel this looks [music] perfect. Now I'll keep the interl like this. Let's try and check if it works. Okay. Now let's click on green flag and check if it works. Uh-oh. There is one problem. As you can see the camera is opened on both the sides. So where will we open [music] the slides then? Right now that is an issue. So I need to switch off this camera [music] first. And I want the backdrop to be visible and not my footage over here. Right. And [music] we just want recognition window to open. So for that what I'm going to do is I will go to machine learning environment [music] and we will put this block over here and I'm going to set the stage transparency to 100% so that the backdrop is properly visible. Okay. Now let's try this once again. It [music] works right now. Let's try the left side first. And that's [music] how my slides are changing just with my gestures. Interesting, isn't it? Now, similarly, you can train Intellio for more creative machine learning projects [music] for object detection or face recognition or pose detection and so much more. Now try it yourself and create some project like maybe currency detector or you can also add the currency in the wallet. [music] Now do let me know your project's name in the comment section. In the next video we will learn four different methods of connection [music] with Intelio and we'll also understand when to use which method for best result. [music] Until then, happy learning.