ক্লাউড ক্যামেরা ট্রাস্ট করবেন না! নিজের সিকিউরিটি নিজেই | Stop Trusting Cloud Cameras

🇧🇩 বাংলা / Bengali / बांग्ला:

NetworkChuck তার দর্শকদের জন্য প্রার্থনা করছেন। তিনি ঈশ্বরের কাছে আশীর্বাদ চাচ্ছেন — আমাদের ক্যারিয়ার, পরিবার এবং জীবনের প্রতিটি ক্ষেত্রে সফলতার জন্য। তিনি বলেন, প্রযুক্তির এই দ্রুত পরিবর্তনশীল যুগে আমরা যেন শান্তি ও দিকনির্দেশনা পাই। তাঁর মতে, আমাদের উচিত পরিবার ও প্রিয়জনদের সাথে সময় কাটানো এবং জীবনের আসল অর্থ খুঁজে বের করা। তিনি যীশুর নামে প্রার্থনা শেষ করেন এবং দর্শকদের জন্য শুভ কামনা করেন।

🇺🇸 English / ইংরেজি / अंग्रेज़ी:

0:00 You need to use frigate right now because right now someone might be watching 0:03 you through your surveillance cameras. Go ahead. 0:05 Look up way high because if your surveillance cameras are connected to the cloud 0:08 in any way, someone might be watching, they sell these feeds on the dark web. 0:13 A threat researcher in Romania told me that and he knows this stuff, 0:16 but with frigate, that’s not going to happen. 0:17 This is a completely local open source AI surveillance system. 0:21 Don’t worry that AI is local and I cannot believe it took me this long to find 0:25 this system. It’s amazing. It has no business being this awesome. 0:28 It’s got facial recognition, license plate recognition, object detection, 0:31 semantic search. It connects to home assistant, 0:34 but the best part is that it doesn’t leave your house. 0:37 It doesn’t touch the internet, 0:38 it doesn’t touch the cloud and sure your ring subscription might say your stuff 0:41 is safe. Do you trust Jeff Bezos really to get you coffee? You’re ready. 0:45 It’s time to take control of your privacy. 0:46 Bring those surveillance cameras in-house. Lock ’em up. Let’s go. 0:54 Now in this video I’m going to show you the quick and dirty setup because it’s 0:57 what I did out of necessity. 0:59 Lemme show you the cool thing about frigate is you can buy whatever security 1:03 cameras you want with an asterisk. We’ll talk about that here in a second. 1:06 But I bought these super cheap cameras, like 56 bucks a piece. 1:09 They’re real link. E one pros and I bought all of these. How awesome is that? 1:13 But I created a weird problem for myself. I’ve deployed so many cameras, 1:16 it destroyed my wireless network. We’ll get to that here in a moment. 1:19 I did solve it I think, 1:21 but the reason I did that is because I was going to Romania to visit. Well, 1:24 Bitdefender and Romania is kind of far and I wanted to keep an eye on my family 1:27 while I was gone. I didn’t have time to run cable. I’m like, 1:29 I’m going to set up a quick surveillance system wireless and I’ll use a 1:32 raspberry pie is my frigate server, which you can totally do. 1:36 And with two cameras, it actually works incredibly well enough talking about it. 1:40 Lemme show you how to do it right now. First, what are you going to need? 1:43 Where’s my pen? We’re going to draw this out. A raspberry pie just yelled at me. 1:47 I’m sorry. The first thing you’ll need is a server. 1:50 Don’t want that word scare you. 1:51 A server can be something as small as a raspberry pi or a spare laptop, 1:54 really anything that can run a docker container because that’s how we’re going 1:57 to install frigate, a docker container on a Linux server. Next, 2:01 you’ll need some cameras. Now pretty much any camera will work, 2:03 but you have to make sure it supports RTSP. 2:06 This is the protocol that will allow it to stream it’s video data to frigate, 2:10 and I hope I’m saying frigate right, because the more I say it, 2:13 the more I feel like I’m not saying it right. It’s okay. 2:15 You’ll correct me in the comments. I know. By the way, I love your comments. 2:18 It’s okay. And have you hacked YouTube today? 2:21 Let’s make sure you do hit that like button subscribe, 2:23 notification bell comment. Let’s hack YouTube today. Ethically, of course. 2:26 Now these cameras again are the real link. E one pros. 2:30 I picked these because they were shippable very quickly from Amazon like next 2:33 day before I left for my trip. I’m a last minute kind of guy, 2:37 but they’re actually kind of amazing. Pan tilt, zoom. 2:40 They don’t need any kind of internet access to function, 2:42 which is important because I’m going to shut ’em down. Cutting you off, buddy. 2:46 I just talked to my camera. I just worked off. 2:49 And that’s the third thing you’re going to need Buffy. 2:53 It’s just required in it or doing anything home lobby or anything privacy 2:57 related. You need coffee. I’m just telling you I didn’t make the rules. 3:00 Never chuck to coffee and back to cameras. 3:02 The brands that they recommend are on a page right here. 3:05 It can be am Crest Dua and even unified Cameras All work. Now, honestly, 3:09 that’s all you really need. Bare minimum. But if you want to make this fun, 3:13 and I know you do, you’re going to want to add this. Where’d it go? Oh, 3:16 it’s right here. For all your Raspberry Pie users, 3:18 which I do recommend a raspberry pie for this if you want to play around, 3:22 they are fun. They have what’s called an AI hat. 3:25 I have the AI hat plus here. 3:27 This sucker is an add-on for your raspberry pie that will help you run AI 3:30 things. Now remember frigate, it uses AI or it can use ai. 3:35 You don’t have to use ai, 3:36 but if you want the cool features like facial recognition, object detection, 3:39 semantic search, this makes it incredibly fast. 3:42 Now I got this from my Raspberry Pi setup, 3:44 but then when I added eight more cameras to it, 3:48 I’m like Raspberry Pi probably can’t handle that. 3:50 So I upgrade to a desktop and I got this thing a Coral USB AI accelerator. 3:55 What’s the actual name? I think it’s a coral TPU. There it is, 3:59 and it’s like a hundred bucks. And this thing is actually kind of amazing. 4:02 I have it plugged into my gaming rig and forget, 4:04 we’ll use that for its AI stuff. That tiny little thing. 4:07 How does it do with that? I don’t know. It’s magic. 4:09 But that’s how we’re keeping our AI local, nothing going to the cloud. 4:12 All right here you can hold the AI stuff in your hand. That’s my motto. Oh, 4:16 I forgot to draw on the extras. Here we go. Bonus AI accelerators. 4:20 I probably didn’t do it right. I didn’t do it right. 4:22 It’s one L and I got to fix it. It’s going to bother me like crazy. 4:26 That feels so much better. Now, again, totally the optional just to get started, 4:29 you don’t need this. Know what this whole project, 4:31 there is one problem I could not solve by myself. You see, 4:33 I was going to Romania another country, right? 4:36 And that was the whole point of this project so I could watch my kids while I’m 4:39 out. Problem is my cell phone wasn’t going to work when I got to Romania. 4:43 Lemme tell you, 4:43 there’s nothing scarier for someone who’s very dependent on their phone. 4:46 I can’t even get around without maps to arrive in a foreign country and not have 4:50 anything to communicate with. 4:52 That’s why I’m extremely grateful for the sponsor of this video. 4:55 Saly before me and my team left, we installed Saly on our phones. Now, 4:59 if you’ve never heard of Saly there, the creators of nor VPN, they’re awesome. 5:02 Now, what’s so cool about this is before I even left, 5:04 I could install Saly and say, you know what? I’m going to Romania. 5:07 I bought myself a Romania pass set up ready to go, 5:10 would not activate until I stepped foot into that country. 5:13 And as soon as we got off that plane, 5:14 I had internet access and throughout the entire trip, now on the way back, 5:18 we actually stopped in Turkey. We got out, saw the sights and yeah, 5:21 you better believe I set up Saily for Turkey before I got there and it totally 5:25 worked. So if you do any kind of traveling, an EIM with Saly is a no brainer. 5:30 I remember the first time I went out of the country, 5:32 it was France and I’m wandering around with a lost puppy trying to find a sim 5:36 card and I finally found one and I’m sitting on the airport floor trying to set 5:39 it up and get it installed. No more of that. Now, 5:42 what was also pretty cool was that nor VPN made Saly. 5:45 So you know Sili has built-in security features. 5:47 You can change your virtual location, 5:49 block malicious ads and stay safer with web protection. 5:52 So if you’re going on a trip anywhere out of the country, 5:54 download Sali right now. 5:55 They cover over 200 destinations and if you scan this QR code right here or just 5:59 use Code network, Chuck at checkout, that’s kind of hard to say. 6:02 Network Chuck at checkout, you’ll get 15% off. 6:05 So scan the code or go to saly.com/network. Chuck, check it out. 6:09 This is my go-to now when traveling and without them I couldn’t have watch my 6:12 kids with frigging. Anyways, back to the video. 6:17 Okay, now it’s time to build our very own AI surveillance system with frigate. 6:20 Now I’m going to demo this with a raspberry pie because it’s stinking small and 6:23 if we can make it work here, it can work on pretty much anything. 6:26 Now we’re going to do this in stages. First, 6:27 we’re going to just do the quickest most base install of frigate on this 6:31 raspberry pie. 6:32 That means we’re not going to use this AI accelerator just yet because not 6:35 everyone’s going to have one and we’ll see how it performs and then we’re going 6:37 to slap this bad boy on there. End up talking. Let’s get going. Now remember, 6:41 we’re going to be deploying this with a docker container, 6:43 which means the steps I’m showing you for the raspberry pie should be applicable 6:46 to pretty much anything because Docker can run almost anywhere. Now, again, 6:50 if I didn’t already say it because I record these in parts, 6:52 get your coffee ready, let’s start building this. 6:54 You can’t have too much coffee, so who cares? 6:56 Now this is a fresh raspberry pie installation. 6:57 I installed raspberry pie os it’s booted up plugging into my network. 7:01 I do recommend a wired connection for the raspberry pie. You don’t need it. 7:04 It can still work. But in this scenario, 7:06 I want to at least one thing wired if we’re going to have these suckers 7:09 wireless. Now, 7:09 I’m not sure if I mentioned before these Rio Link E one pros do have a hardwired 7:13 connection if I want to run cable eventually, but eventually it ain’t right now, 7:18 so I don’t care. So let’s get logged into our raspberry pie. 7:21 Now if you also have a raspberry pie and you’re like, Chuck, 7:22 I’ve never set up a raspberry pie in my entire life, 7:25 I will have instructions below on how to get that set up. 7:27 How to flash your very first image to a raspberry pie and start your raspberry 7:31 pie journey, which is super fun. You can do so much with these things. 7:35 It’s H two network, 7:36 chuck at 10 point 71 70, 7:40 the IP address of my pie. First time. 7:44 Here we go and we’re in now whatever system you’re using, it’s a good practice, 7:48 a good habit just to do a pseudo a PT update, 7:51 updating all your repos that you can install the freshest versions of whatever 7:54 you’re about to do the most fresh. First, we’ll install Docker. 7:57 I’m going to breeze through this because Docker has a great document on their 8:00 website on how to do it. Pretty much just copy and paste, copy and paste this, 8:04 and then copy and paste this command here. Little coffee break while it’s going. 8:09 Alright, Dockers installed. Let’s get frigate set up. First, 8:12 we’ll make a new directory, 8:13 M-K-D-I-R and we’ll call this frigate and then jump in there, CDE frigate. 8:18 Then we’ll create a new file, nano docker, dash compose yml. 8:23 Hit enter. And we’re editing a file. Now we’re going to copy and paste this. 8:26 I will have all this documentation below. 8:28 This is a Docker composed file that describes how to set up a docker 8:31 environment, containers, networking, the works. And if you’re like, Chuck, 8:35 I don’t know what Docker is, I don’t know what compose is. 8:38 I’ve got two videos on that right here. If you want to dive deeper, 8:40 you don’t have to, you can keep going right here with me, 8:42 but you’ll appreciate it more if you go watch those videos. 8:45 Now a couple of things you want to look out for here and you may want to change 8:47 later. We’re mapping some volumes. What does that mean? Well, for example, 8:51 right here, when you’re running frigate, your surveillance system, 8:53 it’s going to be recording things, 8:55 things that your people are doing and it’s got to store that stuff somewhere. 8:58 So this line right here is saying, Hey, 9:00 we’re going to make a new folder called storage on our Raspberry Pi, 9:03 the actual computer, 9:04 but we’re going to map that to a folder inside the docker container, 9:08 the one that frigate will use to store its stuff. 9:11 The same thing goes for config. 9:12 So I point that out because you may want to change where you store your media, 9:16 maybe an external hard drive, maybe a nas, but for now, 9:19 let’s just get this thing up and running. 9:20 One thing you’ll want to change though right now is your password. 9:23 Please don’t leave it at password, please. This might be local and private. 9:27 Don’t make it local and stupid. Now, this is just an example. Don’t do this. 9:30 Now let’s save our stuff. Ctrl X, Y, enter to save. And that’s pretty much it. 9:35 We’re going to spin this up with one command. 9:37 Pseudo docker compose up dash D. Ready, 9:41 set, go. 9:45 Right now it’s going to pull down the frigate image, the docker image, 9:48 little coffee break. 9:49 This might take a minute depending on your internet connection. 9:51 And now it’s creating my image or my container. Done. It’s created, it started. 9:56 Let’s make sure it’s actually working. We’ll do a couple of things real quick. 9:59 I’ll do pseudo docker. PS oh, and look, it actually is running. 10:02 It’s sitting right there. And what you want to look for is the status. 10:05 It says it’s been up for 19 seconds and it’s healthy. 10:07 Sometimes that might change. Let’s run the command one more time. 10:10 It’s still the case now if everything’s working great and you see those 10:12 messages, 10:13 we should be able to open up our web browser and go out to the IP address of our 10:16 pie, which is 10 point 70.7. No, that’s not right. Yes, 10:20 it is one 70 and then port. What was that? 10:23 Port should be 5,000 port 5,000. We did it. 10:27 Nothing’s here, but this is indeed the frigate ui, but there’s no cameras. 10:32 Let’s change that. 10:35 Let’s add our real link cameras or whatever camera you have. Remember, 10:39 your camera has to support RTSP. What is that? 10:42 RTSP is the real time streaming protocol. Is there a dash between real time? 10:46 Let’s go with it. 10:47 IP cameras like Rio Link and a bunch of other different brands, even cheap ones, 10:50 they often have RTSP support. Frigate needs that your camera, 10:54 which this is going to be my camera. It’s my little Rio link. I better label it. 10:59 Your camera’s going to go on your network. 11:01 Let’s just make up an IP address for it real quick. 11:02 If it supports RTSP like ours does, 11:04 frigate will be able to connect to this camera and request a stream. 11:09 And because RTSP is an open standard, it doesn’t matter what camera you get, 11:13 frigate is going to be able to see that stream, bring it in, watch it, 11:16 record it, do some cool AI things with it. And what’s cool about this camera, 11:19 you want to look for this is it supports two streams and mainstream and a 11:22 substream. Why do you need two streams? I’ll show you. 11:25 And it actually helped me solve my wifi issues. Dude, my wifi was so crippled. 11:29 Let’s get reopening set up. Now, 11:31 what I’m about to go through is going to be different depending on what camera 11:33 you have. I’m not going to detail this. 11:35 I’m going to walk through it very quickly so you can kind of be familiar with 11:37 what you’re going to look for. Now, 11:39 I’m just going to plug these into my power adapter. Be right back. 11:42 And one thing very annoying about this, oh, there it goes. It’s going on. 11:45 Did I mention these things? Your PTV pen, tilt, zoom. That’s so cool. 11:49 And you’ll be able to use frigate to control that too. Welcome to link. 11:53 Please install real link app and scale the QR code on the camera. 11:58 Comment by, 12:12 okay, it’s done. Now one annoying thing about this is that most cameras, 12:15 they do have a web interface. So if you can find the IP address, 12:18 once you connect this camera, you can just connect to the web interface, 12:21 do your stuff. This thing does not have a web interface at boot by default, 12:26 so I’m going to have to use the Rio Link app, 12:28 but only really one time just for setup and config changes. 12:31 But it’s not going out to the big bad wild internet. 12:33 It’s staying local right here. So I’ll go add it, click on the add button. 12:36 I’ll scan that little QR code we have here. There he is. 12:40 And I’ll configure it via Bluetooth. 12:45 Connect it to wifi, give it a password, name it, 12:50 and that’s it. Look, it’s working. Oh wait, hold on. 12:56 Now by default it’s not going to have RTSP enabled. 12:58 Let’s go enable it right now. Again, 12:59 these are some of the things you’ll look for when you configure any camera. 13:02 I’ll go to settings here. 13:03 I’ll click on the device to go to more device info network information. 13:07 They kind of hide these settings advanced. 13:10 Now we’re here down here at the bottom. We got RTSP. I’ll enable it. Yes, I do. 13:15 The default port is fine. And then O-N-V-I-F. 13:18 That’s for controlling it. That’s a whole separate protocol. 13:22 And then that’s it. I think it already saved it for me. 13:26 Notice I’m not logged into the cloud. I don’t have to be. 13:28 It’s a direct connection from my phone to the camera’s IP address. 13:31 We’re on the same network. And before we try and connect it to frigate, 13:34 let’s make sure the RTSP stream actually works. We can do that via command line. 13:37 It’s kind of fun. First we’ll make sure we have FM PEG installed. 13:40 So we’ll do a pseudo A PT install FM peg. 13:43 This is a great open source project that everyone uses for a lot of things. 13:48 You can even use it to edit videos. 13:49 And notice I’m doing this on my pie because I’m going to test the connectivity 13:52 between the camera and the pie before we try and add it to frigate. 13:56 Doing that little step will help us avoid any troubleshooting rabbit holes. 14:00 We might go down, it’s almost like I’ve been there before. 14:03 Are we’re going to use this command with FFM peg. 14:05 And this is important because you’re going to see how the R-T-S-P-U-R-L looks. 14:08 We’re going to use the same URL to configure frigate. So ffm PEG dash i. 14:13 Then we have the R-T-S-P-U-R-L. It starts with our username and password. 14:17 Now my password’s going to be blurred out, right Video editor. So username, 14:22 password. Then we have our host name, the IP address, the port number, 14:26 which we just saw on our app. 14:27 And then a forward slash with the location of our stream or the certain stream. 14:31 We’re actually looking at that this camera has configured. 14:33 Now this might be different depending on your camera. 14:36 So look at the documentation. 14:37 But this format for the R-T-S-P-U-R-L will be the same across the board. 14:42 Let’s test it out. Oh, I forgot the rest of it. I got a backslash. 14:45 I can continue the command on the next line here I’m saying, hey, 14:48 let’s create a video. 14:50 Stream it from this camera for five seconds and let’s copy it to a file name 14:53 test MP four. Let’s try it out. Ready, set, go. 14:57 It’s going to think about it and there it’s actually pulling in stuff and it did 15:00 it. Let’s see if file was made. I’ll type in ls. There it is. Test MP four. 15:03 I’m going to pull that over to my desktop to take a look at it. I’ll use SEP. 15:07 Got it. Let’s see how it looks. 15:12 It worked. Now let’s add this camera to frigate. 15:14 So first we’ll go into our config folder, 15:17 which if we type in LS right here in our frigate directory, 15:19 we should have a new config folder that was created by Docker. 15:23 We’ll jump in there, CD config, and then we’ll create a new file with Nano. 15:27 So we’ll do pseudo nano config y AML yaml. 15:31 Now notice they already had kind of a blank config for us for the camera. 15:34 Just a test ip. Alright, so we’re going to cut out the base config. 15:37 We’re going to add this config. 15:38 Now there’s a few things we have going on that are different. 15:40 First we have an input argument. 15:42 They have some presets on how to ingest the RTSP stream. Easier for me to say, 15:46 this one’s really good for Rio Link cameras. 15:48 You can check the website documentation on, forget to see what they got. 15:52 Then here I’ve added my camera. Notice the URL. Oh wait, the IP address. 15:55 Need to fix that. It’s very similar to what we use to test with F feg. 15:59 It almost identical because it is identical. And then just for testing, 16:02 we do have the role of this stream detect, but we have it disabled. 16:06 We have enabled false just for testing to make sure things work. So here we go. 16:10 Control xy, enter to save. 16:13 Let’s bring down our container because right now for it gets not using that new 16:16 config. And let’s bring you back up. Pseudo ducker, compose up dash D. 16:21 Ready, set, go. Start it. Sorry. Created, started. Let’s do a docker. 16:26 PS it’s starting. It’s starting. 16:30 Starting up healthy. Cool. Let’s go see if it’s working. Ah, 16:35 now it might take a second to come up for you, but dude, we have contact. 16:40 Now it might be like, Chuck, that looks really bad. 16:42 That’s because we’re using the substream. We’ll get to that here in a second. 16:45 Dude, look at you. That’s what you look like to me. 16:49 That’s going to be my new keyboard camera. Now at this point, 16:53 let’s take a snapshot and get our bearings. At the bottom left, 16:55 you’ll see CPU usage. It’s not really doing much. 16:58 So 5% it’s average. That’s good system. 17:01 Over here on the bottom right is super healthy. But look at what we’ve done. 17:05 We have a camera added to frigate and it wasn’t that hard. 17:08 Now there might be some troubleshooting. I get that. 17:10 Refer to your documentation of your camera to see what the RTSP or LS are and 17:14 everything. But if you hold along with that config, it should work. 17:20 Now what do you say? We actually do some detection. Let’s make it happen. 17:23 This’ll be using ai, so it’s going to be heavy on this poor little pie. 17:26 Let’s make a Soweto. Let’s edit our config once more using nano. 17:29 And over here on detect, 17:30 we’re going to change it from enabled false to my voice. 17:33 Kind of cracked there to enabled true. And then just above our camera section, 17:37 I’m going to add another section called objects. Just like this. Hit a colon, 17:41 enter tab over and we’ll say track colon, enter tab. 17:46 Tab over. We’ll do a dash and we’ll say person and then keyboard. 17:50 They have a whole number of things you can track. I think they track these. 17:53 Definitely a person. I’m not sure about keyboard. 17:55 But what us telling it to do is actually look for these things and let you know 17:58 when it sees it. That should be all we need for that config. Go hit control. X, 18:01 Y. Enter to save. Bring our container down. We could do a restart. 18:04 I prefer to do down than up just to clean. I like to watch it happen. 18:08 The one do up. It’s up. Let’s do a ps. It’s still starting. 18:14 It’s up and healthy. Let’s go check it out. Alright, 18:16 start it up and already recognizes the keyboard. Look right here. Keyboard. 18:22 That’s awesome. Notices how the camera’s wrapped around and red or the feed is, 18:25 let’s see how it does with people. I’m going to refresh my page. 18:30 So you can see the CPU spike. Where’s the CPU? Give me my CPU stats buddy. 18:35 We’ll give it a second. Let it catch up to itself there. It’s CPU right there. 18:38 Look at it. It’s 41%. Trying to track that keyboard. 18:41 Let’s see how it does with people. Aha person, that’s awesome. 18:45 But 41% with one camera. We’re not even recording, we’re just detecting. 18:49 It’s now to enable recording. And we’ll do the recording based on motion. 18:53 You can actually do it based on a lot of things. We’ll get back into our config, 18:56 pseudo nano config and under inputs. We’re going to add one more. 18:59 It’ll be a new path option. And if your camera supports two feeds like mine, 19:03 I’m not going to do the sub feed. I’m going to do the main feed. 19:06 Now what does that mean? Well, right here we have our sub feed, 19:09 which is actually a lower quality feed down here. 19:12 We have a bit more of that config. The video itself is a smaller size. 19:15 Frames per second is five. Very small, 19:17 just enough for the AI itself to see things and detect things. This feed, 19:22 the main feed is high quality. It’s for recording, 19:24 it’s for people review for us to see it. 19:27 And we enable that by giving it the rolls record. 19:30 So now we have two feeds going to our Raspberry Pi, our frigate, 19:33 we have the detect feed and we have a record feed. 19:36 These are two separate RTSP streams being streamed from the camera to the 19:40 Raspberry Pi. Now I had 10 cameras. 19:42 Now you’re understanding why the wifi might’ve been like blah. 19:44 Now there’s one more thing we have to add here. Notice we do have detect config, 19:47 but we’ll also need record config. 19:50 So we’ll actually add this right underneath our objects config and I’ll have all 19:53 commands below, all config below. You can copy and paste. 19:56 And this config is pretty basic enabling recording for seven days. 20:00 And the mode is motion only record if you see things moving. Okay, 20:03 the one do control X, Y enter to save. 20:05 We’ll bring our stuff down and then back up docker PS to see if it actually 20:09 comes up. Okay, it’s healthy. Alright, let’s go check it out. Refresh our page. 20:13 Okay, so actually you see at the top there, we already have some recordings. 20:17 Motion. Hello? Another recording. So if we go to review, 20:22 there’s a recording, let’s play it. That’s pretty sick. 20:26 Then we have our detections tab. It actually show like dude keyboard. 20:30 If we go to our search tab right here, 20:32 we can see the keyboard or person detections. Jump in there, go to video, 20:37 watch that. That’s so cool. Person. Where was a person detected here? Seriously? 20:42 Where was a person detected? Oh, right there when I turned up to my face. 20:46 You see how powerful this is? This is your very own surveillance system. 20:51 Now CPU is at 26%. Let’s go check out the average. 20:55 We can go to settings and then system metrics. Oh, detector CPU usage, 20:59 206%. This thing’s being hit like crazy. 21:03 So these are metrics we want to pay attention to. 21:05 We’re about to do some fun stuff. But CPU detection speed, 21:08 61.93 milliseconds, which is fine. And the CPU usage, 21:13 which is not fine at all. Now you’re seeing where things hurt. 21:16 I’m going to add a second camera and see how badly we can make you cry. 21:19 Give me a second. By the way, 21:20 the way it’s able to detect objects is with yolo. I think it’s using yolo. 21:25 V eight CPU is slow. Yeah, no kidding buddy. 21:29 Welcome to Rio Link. 21:35 Now for my second camera, I’m just going to repeat this config naming it. 21:39 Rio Link two and changing the IP address for each path. It’s simple. 21:45 Okay, it’s up. Let’s refresh our page. 21:48 I think the first camera or the second camera is still initializing. 21:50 You might see that when it’s first getting started up and doing. Oh, yep, 21:54 there it goes. Okay, so now we have two cameras. Both are doing detection, 21:58 both are recording. Here we go. Me two angles, 22:02 me and my keyboard. And that should be really hurting this respiratory body, 22:07 but also it’s working, which is pretty impressive. 22:09 So now we’re at like 82.26 milliseconds on detector inference speed. 22:14 CPU is still just being murdered. It’s telling us CPU is slow. Yeah, we know. 22:19 What do you say? We give him some help. Real quick, 22:23 I want to see what AI hardware does for this guy. Now for this, 22:26 I’m going to shut down my pie and I’m going to install this on top. It’s a hat. 22:30 He’s going to wear it. Now I’m just kind of doing this in a janky way. 22:34 I don’t want to remove anything. But there it is, 22:38 connected to the PCIE slot on the Raspberry PI five. 22:41 We’ll get ’em powered back up. Now with the halo setup, 22:43 I’m going to install a few drivers, change out the docker config. 22:46 I’m not going to bore you with the details. 22:47 I will have configs down below if you want to have this exact setup. 22:50 In full disclosure, I’ve recorded this twice. 22:52 Now you’re seeing the second version of me where I troubleshooted the crap out 22:55 of this because the way I did it before three or four months ago has changed, 22:59 but I got it working now. And check this out. Woo. 23:03 And look at the CPU 14%. If we go to our settings here, 23:07 we go to system metrics. Notice a few things. 23:11 And you see the power of this right now. We now have halo. 23:16 It’s only at 4.1%. Look at the inference speed. 23:20 Before it was like 80 something, now it’s at nine. Are you kidding me? 23:25 That’s amazing. That’s amazing. Hold on, we’re not even looking at it right now. 23:30 That’s amazing. Now let’s detect a keyboard and we’re just like, 23:34 we’re doing all this crazy stuff and the metrics are great. 23:37 AI accelerators are awesome. Now we could also enable some other settings too. 23:41 If I go to settings and I go to enrichments, enrichments, 23:46 I love that word, we can do things like semantic search. 23:49 We can search for objects that are being detected and choose what model to run 23:52 small or large. We can do face recognition, license plate, bird. 23:57 I can run all this stuff and then restart that to apply to the config. 24:00 Let’s do that. 24:01 So right now I’ve got two IP cameras streaming to my Raspberry Pi 24:06 with an AI accelerator. It’s recording, it’s doing motion detection, 24:09 object detection with ai. It’s got all the AI features enabled semantic search. 24:14 And it’s not even being hit that hard with this halo hat with the device, 24:17 this small being able to run your own surveillance. That’s crazy. 24:20 Now I forgot the prices. 24:20 I’ll have the prices right here when the video editors add it. 24:23 But this isn’t too bad to have a self-hosted surveillance system. 24:26 And you know what I forgot to add on V. 24:28 I want to be able to control these cameras. Let’s make that change real quick. 24:31 I’ll just add this config to one camera here. Let’s check it out. Okay, 24:35 it’s back up. Let’s go to the camera. Ah, we have controls now check it out. 24:39 Here we go. 24:43 That’s awesome. That’s so cool, 24:48 right? And the best thing ever is that I’m controlling this. It’s in my control. 24:52 No one depends. It’s mine. You know what I mean? 24:56 It’s exciting to own your own stuff. Own your own everything. 25:00 Internet goes down. I don’t care. It’s all local. I don’t have Zoom. 25:03 I was trying to do Zoom. This is so fun. 25:08 Let’s go look at my face detection. Look at that. 25:12 So I added a photo of me by going to add face and it recognized that it’s hey, 25:16 it’s me and I can train it. So yeah, that’s me. Yeah, that’s me. 25:20 So it gets better at recognizing me. And I did that with everyone in my house. 25:23 You’ll see that later. But here’s the thing, 25:25 adding eight more cameras I think is going to kill this thing, 25:30 which is why I decided to move up to a bigger machine, a desktop. 25:36 This is actually one of the PCs I used in my daughter PC bill back in the day. 25:40 They don’t use ’em anymore and it’s just been collecting dust, 25:43 but I’m putting them back in it’s time for frigate. Now this guy’s pretty good. 25:46 He has a rise in CPU and Nvidia 2070 GPU. And then this little guy right here, 25:51 he’s, 25:51 even though this has a killer CPU and a great GPU for what we’re about to do, 25:56 it still needed AI acceleration. 25:58 Now I couldn’t add this halo hat to it and I was impatient because I was about 26:02 to leave for Romania. So I’m like, I need something quick. Two day shipping, 26:05 let’s go. And I found the Google Coral USB accelerator similar to the halo. 26:09 This thing’s designed to do AI stuff. That’s all it cares about. 26:12 But it comes on the USB form factor and I can plug it into any computer and use 26:15 it. Now here’s how things work. The Nvidia GPU, it’s being used. 26:19 It’s being used for video decoding. It takes those 10 RTSP streams. 26:23 Or if you’re doing substreams, 26:24 it takes 20 and turns ’em into frames that frigate can use. Coral. 26:28 He has one job. I’m going to call him Carl. Carl has one job, AI inference. 26:33 It’s crazy. This little dongle hanging off. 26:36 My computer is handling all of that. 26:38 So I rapidly install these guys because I’m leaving the next day. 26:41 10 reel link cameras. Five downstairs, five upstairs. And remember I’m lazy. 26:45 I’m not going to go and install or run cable everywhere in a two story house. 26:50 No thank you. I’ll make wifi work mistake. 26:55 I thought it was going to be easy. No, but dude, 26:57 I got a set up and you know what? It was working great. Look at that. 27:01 10 cameras. Look at my utilization. 27:04 I got my Nvidia super locked in CPU 27%. If I jump into my system, 27:08 metrics detector, CPU usage, it’s high. It’s my coral. 27:12 You can see the coral right there. 27:13 Inference speed is 10 milliseconds for 10 cameras. That’s solid. 27:17 And they look at my face detection. It’s fun. 27:19 All types of faces coming up in here. That’s not Millie, 27:23 that’s Addie boom. It’s kind of like Google photos. He can match the faces. 27:27 You go in there every once in a while, match up some faces. 27:30 The problem is my kids all look like my wife, so they’re all my wife. 27:34 I can search for things barking. 27:38 What’s happening here? 27:45 It sounds like a horror movie. That’s a house with six kids. Where are my kids? 27:50 Others? One watching tv. So what you just saw there, 27:55 it worked great for about 12 hours. 12 hours, didn’t slight. 27:58 My network got haunted. I’d be sitting there working and no pages loading. 28:02 And then I look over at Frigate. Dude, a scene out of Pulter guys, 28:06 just things flickering, cameras going nuts. Nothing loading. 28:10 Like what the heck is happening? 28:11 I had to take my morning standup with my team on my 5G connection. 28:14 Things are just dead. That’s embarrassing for network, Chuck. 28:17 I totally get that. So troubleshooting, I’m like, okay, what changed my network? 28:21 Well, I added frigate. So lemme shut frigate down. Boom, wifi comes back up. 28:25 Okay, lemme bring frigate back up. Things are working 12 hours later. Boom. 28:30 It blows up again. What is happening actually in the comments right now, 28:34 let me know what do you think is happening? And it’s not what you think. 28:38 It’s not what I thought. But I didn’t have time. I had to take up to Romania. 28:40 So I left. And when I tried to log back in and watch my house and my cameras, 28:43 it was all messed up again. I restarted the docker container, 28:46 it’d be fine again. And I had to do this pretty often. And when I got back, 28:49 I was just done. I shut down for a again, I said, I’m writing you off. 28:53 Video canceled. I was so excited to talk about this within last week. I decided, 28:57 you know what? I’m going to figure this stinking out. 28:59 Forget is too much of a cool idea. It’s got to be a me issue. 29:02 Let’s put the network back in, network Chuck and troubleshoot this. 29:05 And I know what you’re thinking. You’re like, Chuck, it’s bandwidth. 29:07 You’re destroying your wifi. There’s too many things. Hey, wasn’t, 29:11 get your coffee ready. We’re about to get a little nerdy with this. 29:15 I’ll see you in the next segment. 29:18 So I abandoned the project for two months, but I came back with determination. 29:22 At first I thought it was bandwidth. That’s obvious, right? 29:24 You add 10 cameras that are streaming, two streams, 29:27 a video on your wifi, something’s got to be hit, right? No, 29:31 I jump into my unified controller, which is what I use for my network. 29:35 Bandwidth was not even touched, barely anything. This is all my local network. 29:40 And I kind of had that assumption, right? It worked for 12 hours. 29:43 What’s happening at that 12 hour mark? 29:45 And why is it working great for those 12 hours? That doesn’t make any sense. 29:48 But then I dug deeper. I jumped into my unified controller. 29:51 I’m looking at all the wifi information, I’m like, man, 29:53 there’s got to be something with this. And I saw this metric on Dumble door. 29:58 Dumble door is my AP upstairs. I got two aps in my house. 30:00 One downstairs is Hagrid, one upstairs is Dumble door. 30:03 And I saw this metric TX retry rate. Let’s get to some drawing. 30:07 So I saw this metric on Dumble door called the TX retry rate. 30:10 Dumble doors was really high. 30:13 And around 12 hours after I rebooted frigate, it was around 29%. 30:18 Now what does that mean? Well, 30:19 when Dumble door tries to send a packet to the IP camera, 30:22 if it gets there and the IP camera acknowledges that success, 30:25 if it doesn’t acknowledge it, then it’s going to retry it. 30:28 It’s going to try and rescind it. And that’s the retry rate at 29%. 30:32 That means three out of every packets failed on the first transmit, 30:36 which is bad. If you look at my AP downstairs, Hagrid, he’s a 7%. 30:41 That’s respectable in the wifi world. So that told me something. 30:43 It wasn’t a bandwidth issue, it was an air issue. 30:47 I had some wifi saturation. It’s kind of like this. 30:49 Imagine 15 people are all trying to talk together, 30:52 having different conversations and a room that only fits five people, 30:55 everyone’s talking over each other, they can’t hear each other. 30:58 They’re constantly repeating like, oh, what’d you say? Oh, what’d you say? 31:01 Then imagine that just happening for 24 hours straight. 31:03 That’s kind of what 10 wireless cameras were doing to my wifi. 31:06 Now this makes sense in theory, 31:08 but why after 12 hours and it was sometimes like 20 hours, it would vary. 31:12 Well, 31:13 because these cameras would be sending multiple streams and because there’s so 31:16 many cameras transmitting so many packets, 31:18 they start to fail in ap and my access point will. 31:23 And over time, more and more retransmit are happening. 31:26 And the problem is that the connection between the camera and frit is 31:28 technically up. But packets are failing constantly. 31:31 And over time the memory gets full, the buffer gets full. It’s like the Titanic. 31:36 It just over time filled up and it’s gone. 31:39 But then when I reboot Frigate Fresh Connection, 31:41 essentially it was like rush hour traffic except the cars kept failing to go 31:44 home. So they added more cars to try and retry. And before you know it, 31:48 there’s just so many cars on the highway, nothing moved. 31:50 And it wasn’t just that I found out also that Rio Link cameras, 31:54 specifically my model, the E one Pro, it’s known for having RTSP issues, 31:58 specifically their stream. 32:00 When it goes roughly over 20 hours starts to degrade in quality could freeze. 32:05 And I’m like, oh my gosh, that’s my issue. So there’s another thing. 32:08 And what’s funny is real Link even has an option in their camera 32:13 config to reboot on a schedule. And at first when I saw that setting, 32:16 cause I poured through settings, I’m a nerd and I’m like, 32:19 why do they have a reboot schedule? Why You can do it every week, every day. 32:23 And now I get it, 32:23 they want you to reboot their cameras periodically because the RTSP issue is 32:27 there. So I did three things to fix my problem here. 32:31 Problems. First I optimized some configuration. Second, I automated some resets, 32:35 real link. And three, I added another access point. 32:39 First in Rio Link I enabled a thing called CBR or constant bit rate. 32:43 They call it fluency first in Rio Link. 32:45 Some IP cameras like Rio Link try to have variable bit rate to handle bandwidth 32:49 fluctuations. But that made it a bit unpredictable. For my wifi network, 32:52 I wanted a predictable bit rate, which helps wifi handle it better. 32:56 I also made sure I was using TCP instead of UDP. Rio Link does better with TCP. 33:00 TCP is more stable. Uud P is lossy doesn’t really care if packets get there. 33:04 Three, I use a technology called Go to RTC. 33:07 This is actually pretty awesome because in my environment I have my frigate 33:09 server, which was receiving a stream from my IP cameras. 33:12 And then when I used integrations like home assistant, 33:14 which is awesome by the way, it would also receive a stream from that camera. 33:18 Or when I use my phone to watch a stream, 33:21 even if I’m using a frigates web interface, 33:23 I would still receive the URL directly connecting to that camera. 33:26 What go RTC does? It says, no, no, no, no, none of this. 33:29 Connecting to the camera multiple times from multiple places, 33:31 we’re going to have one connection to Frigate, 33:34 it handles that and then it’ll kind of advertise that stream from local host. 33:38 So then every other device will now connect directly to Frigate for that stream, 33:42 which frigate, my server is hardwired, here’s what that config looks like. 33:45 It’s a bit weird. So I’ll have all that documentation below. 33:47 Frigate can be very complex. 33:50 It can be as simple as our first example in the beginning with our raspberry 33:53 pie. You can go crazy and going crazy is pretty fun. 33:56 See here I have a section called Go RTC. And this receives all the streams. 34:00 And then the actual camera section is local host fix number two is a no brainer. 34:04 If a reboot fixes the cameras, then reboot them on a schedule. 34:07 So in each Relink camera I enable to reboot, I’ll limit it to once per 24 hours. 34:12 So I staggered them over that time. 34:14 Kitchen reboot at 3:00 AM backyard reboot at 12:00 PM The only downside is I’m 34:18 missing like what? 34:19 Three seconds of footage If someone wants to time something and comes still 34:24 a inflatable dog for my backyard or something, I don’t know. 34:27 Why did I say inflatable dog? I don’t have that. 34:30 And then this third solution was adding another access point that airtime was 34:33 getting crazy. I wanted to clean that up a bit. 34:35 So instead of having five cameras on one AP and five cameras on another, 34:38 I added a third ap, which is about three to four cameras per ap. 34:41 And after I added those three fixes, I’ve had no issues with frigate at all. 34:46 It’s been running smooth for like a week and a half. 34:49 And lemme show you the network. Actually I haven’t looked at this in a while. 34:53 I’m really curious. Specifically I want to see the retry rates. 34:56 So this is Dumbledore for the past 24 hours. He’s doing like what, 34:59 7% on average, 6%. 35:01 I noticed I’m looking specifically at five gigahertz on wifi. 35:04 I got two options for Spectrum 2.4 or five. 35:07 I chose five because five gives you more bandwidth. And I’m streaming video. 35:10 You’re probably wondering, Chuck, how did you only do five? Well, in Unify, 35:13 and you might have this on most access points, when I configure my wifi network, 35:17 they call it saron. The all seeing I, I’m only advertising five gigahertz. 35:20 So every camera is connecting via five gigahertz. 35:23 I’ve locked the AP they connect to. So when all the aps are up, 35:26 they have one they report to. I also assigned them static ip. 35:29 So I don’t have any D-H-C-P-G craziness going on. 35:31 Let’s check the other aps like my new one. How’s he doing? Whoa, 35:34 his are actually pretty high. What’s happening? 35:37 So it didn’t hit the max of like 29%, but it’s still not doing okay, 35:40 we’re discovering this together. 35:41 Let’s go see what clients are connected to him on so on. 35:44 It only has two connections right now. That’s so strange. 35:48 So maybe the story’s not over, but the reboots are sure working. Now, 35:51 you could avoid all these issues by just running ethernet and you should do 35:54 that. But if you want to make it work with wifi, 35:57 if you live in an apartment where you can’t punch holes in the wall where you’re 35:59 lazy like me, you can make wifi work. 36:01 It’s probably easy and a no brainer if you’re running three to five cameras. 36:04 If you’re getting up to 10, you got to optimize a few things. 36:06 You just go to RTC, 36:08 make sure you got enough access points to deal with all the airtime, 36:10 the air traffic, 36:11 and sometimes you have real link cameras that just need to be rebooted on a 36:14 schedule 80, 55 bucks. I’m okay with that reality because at the end of the day, 36:18 I have local private surveillance with ai, 36:23 no cloud of seeing my stuff, no one’s peering into my house. These cameras, 36:27 they don’t touch the internet. 36:28 I can have them blocked off and also I can connect them to a lot of cool things 36:32 like home assistant. Now my wife is not going to log into Frigate, 36:35 she’s not going to do it, but she does already use Home Assistant. 36:37 I’m not going to show you how to use this integration. It’s a whole thing. 36:40 That could be another video. But once you add it in, you got some awesome stuff. 36:44 It’ll show you when it finds people or umbrellas, 36:49 which is not an umbrella, you can jump into the feeds. 36:51 Looks like my wife is making dinner. I need to hurry up. 36:54 And I’m looking at one of the cameras here. 36:55 Look at all the sensors it has and all the automations you could do with this 36:59 stuff. It’s so powerful. So let’s forget. 37:01 Lemme know what you think in the comments below. Are you going to try this? 37:03 Do you already have it? Can you teach me something? 37:06 Have you done something cool I don’t even know about? I would love to hear that. 37:09 Anyways, that’s all I got. I think this video is probably a bit long. 37:11 Hopefully you get a chance to deploy this and hopefully I helped you along with 37:15 a few of the issues that I ran into because this took me a minute. 37:19 That’s all I got. I’ll catch you guys next time. Hey, 37:22 you made it to the end of the video. You’re still here. 37:25 I just started recently and by recently, I mean in the past month or so, 37:29 praying for you. I’m a believer, 37:32 I am a Christian and I know many of you aren’t and that’s fine. 37:36 I would love to pray for you. 37:38 I know it’s kind of weird and it may not be your thing. That’s cool. 37:40 You can click off. But if you want to stay and stick around, 37:43 I want to pray for your life, your family. Yeah, that’s it. Let’s go. God, 37:47 I thank you for the person watching this video. 37:49 I thank you that they are passionate about it. 37:53 I thank you that they value privacy and I ask that you bless them now, 37:57 bless them as they deploy this technology or as they’re considering it, 38:02 let it go. Well let it be successful. But ultimately, 38:04 I pray for their lives that they would be filled with joy, 38:07 that they’d be filled with peace, that their families would be blessed, 38:10 that they would feel satisfied in their career and that their career kind of 38:15 sucks right now. God, just give ’em a path forward. Let them, 38:19 if they’re trying to jump into it as a career, Lord, just bless them in that. 38:22 Give them discipline and let them be steadfast with their study and 38:27 give them great favor as they’re applying for jobs. 38:29 Let them find that perfect opportunity. God and Lord, 38:32 I ask if they’re going through anything hard right now, 38:34 if they’re dealing with just things that are bogging them down, frustrations, 38:39 family issues, depression, I ask that you just release that from them. God, 38:43 that you remind them of who they are in you. 38:46 I remind them of the good things they have in their life and just fill them with 38:49 joy today. I pray blessings over their families, over their lives, 38:53 over their careers. God, just give them what they need right now. 38:58 I pray this in Your name, Jesus. Amen. 39:01 Thank you for letting me do that and I’m extremely grateful for you. 39:04 And the reason I want to pray for you is I love my audience. I love you, 39:07 and I truly believe that prayer works. 39:10 And while I can share all the knowledge I can and make videos that I think 39:14 you’ll enjoy, I think this part I believe will have the biggest impact on you. 39:18 I hope so. That’s why I do it. Yeah, 39:22 I’ll catch you guys next time.

🇮🇳 हिन्दी / Hindi / হিন্দি:

NetworkChuck अपने दर्शकों के लिए प्रार्थना कर रहे हैं। वह ईश्वर से आशीर्वाद माँग रहे हैं — हमारे करियर, परिवार और जीवन के हर क्षेत्र में सफलता के लिए। वह कहते हैं कि तकनीक के इस तेज़ी से बदलते दौर में हम शांति और दिशा पाएँ। उनके अनुसार, हमें परिवार और प्रियजनों के साथ समय बिताना चाहिए और जीवन का सही अर्थ खोजना चाहिए। वह यीशु के नाम में प्रार्थना समाप्त करते हैं।