Nural Network Aimbot Tutorial

all right ladies and gentlemen today i'm presenting you something quite amazing today we're gonna be learning how to do our own neural network aimbot now this code isn't mine uh yeah i'm just going to teach you how you will go about and just make your own model for any game because i'm the code i'm going to use is specifically trained for a game i'm going to teach you how to use it across any game all right so first thing we're going to want to do pull up our uh internet explorer of choice i'm just too lazy to install google so first thing we want to install is auto hot key all right download i'm gonna download current version quick download doesn't take any time you press yes on the thing uh do we wanna we wanna do custom installation ansi 32-bit uh program yep keep that default we're good and exit all right the next thing we are going to want to do now all of this stuff that i will be pasting in the video will be in the video description so uh go to new create auto hotkey script title this i don't know shot that's it we're good now i'm gonna because i've uh i've emailed myself i need to do right here take that it's gonna be into never trust control c i always got to do that manual copying all right okay it's gonna be in the video description just paste it all right auto hotkey script edit script no one needs this who cares uh do that ctrl v all right let's talk about what the script does uh ctrl s to save remember very important or you can go to file and then save so it's very important what does this do this takes screenshots for us f5 is the key that takes screenshots in steam this is the screenshots we will be using to train our custom models what is let's let's analyze this script right here p to start click p that will start taking screenshots to pause the script hit the pause slash break button on your keyboard and then to start it again hit the pause button and then to uh and then to pause it hit the pause so basically it just unstarts and restarts that's what this does so very important what you're going to want to do whenever you see an enemy or whenever you're like you know and a place where you're ever gonna fight really turn this script on press p or unpause it by pause break okay you don't have to do this but somehow you have to get a bunch of images of players it's personally my recommended way so once you do this you want to double click or you want to type in here and you want to type screen screen shot and you want to enter now if you ran it again you'll see this instance just click yes no big deal but all right so press p it'll start taking pictures by pressing f5 which is how you take pictures in steam any steam game all right so let's um let's let's collect some clips really i'm gonna boot up cs go and collect some clips i'm gonna go over exactly where the clips are stored where to upload them how to train them all right guys see you in a bit just remember collect those clips you want around a thousand that is around the baseline i'm gonna be training with around 500 because i'm lazy it's my chronic illness laziness but you will want a thousand images minimum preferably the more the better you cannot have not enough images this other thing you can have a million images the more the better all right so 1000 images shoot for that goal alright see you back in a minute so i i decided to do a little bit of the [Music] of recording while i'm playing my game so right now i'm playing uh against bots in cs go i'm playing the gun game because it's by far the easiest game mode as you can see run every second in that very very annoying screenshot noise you do have to deal with this um i'm sorry to say but as you can see it's i mean it's pretty effective we are lagging quite a bit see this is some pro cs go move and that learned that epic flicking from surf you know so nice things whenever you die this thing sometimes stops but if if i want to stop let's sit with pause break that's it stops no more noise pause break again boom it starts whining um so i mean yeah this is pretty much it um remember from the thousand images all right we are back after you know uh you know setting it up collecting data showing you how to do that so right now we need to see how much data we collected and to do that we have to know where it's stored on our disk when we take screenshots from steam so let's um all right so go to the game i mean i'm using a um i don't know alt pc it's not really my main piece here's why i only have one steam game so many screenshots be brought to this window right here click show on disk click that all right you'll see we have tons of screenshots but we don't know how many to do that we would go here right so we were right here we'd click on this 730 might be a little different for you click on this click on properties 800 files now only the files with player detections will be used in our training only files with player detections will be used in our training all right only those files no other files will be used so all the files without players it's not going to help you so i probably divide this number well i mean depends on the game but you you know for me two-thirds around 500. for you if you're just walking around and not doing much you get one or two minutes of combat not going to cut it you're not yeah and remember that's why you can pause the script only activate it when there's combat all right so we have all of our images here next what are we going to do we're going to want to annotate our images now to do that to go back to my all handy thingy thingymabobber right we're going to want to do google this anaconda python oh that's not how you spell python anaconda python go here download all right we're gonna wait for it to download all right after a minute our download has finished um open file obviously you can do this from the downloads folder because that's where your downloads go if you don't understand that probably too too advanced of a tutorial for you so um don't add it to your path i mean just keep all the settings default really there's no problem okay this is going to be done a few seconds yeah all right just stuck um [Music] okay i assume i don't really feel like editing any of this out like i'm far too lazy to do this i don't even think i'm gonna edit out the opening and closing of obs and right now i'm just waiting for it to download i toggled it but that was about it right you know this was just so really a pain in the butt right now um it says done but it's clearly not kind of annoying me but i mean while we're waiting for this let's go back and let's see what we're doing we're gonna install label img really easy to select players and not really players or whatever you want that's the first thing that came to mind we're going to be using anaconda this is just what we installed installed anaconda here um so ah here we go ctrl u load all of the images from a directory r control s save copy the current label in 250 delete current image space flag of the current image as verified w create rectangle d next image a previous dell delete the rectangle and ct rl plus plus zoom and zoom out um pretty we're only going to be using mainly a adw and ctrl s to save i'm showing you how to set that up that's actually why we installed anaconda great looks like this is done finish we're all it opens two tabs very annoying please don't do that mr developers um so what i didn't know why he showed up there but um looking at me um all right remember all this is in the description so we copy just oh remember we opened anaconda in here just type in anaconda you'll see the anaconda cnd paste the first line in wait a little bit yes done all right we're good paste the second thing in lxml yes so you don't have to do this this is just preferably what i want to do if you really want you can skip to the video where we upload roboflow and you can it's called annotating your you can annotate your images now we're pasting this oh who cares all right this should work i really hope please work um right we're morons because i totally forgot to to code download it's a real big brand of me um yeah download boom extract all boom all right really stupid that is me so here's what we're actually going to want to do all of those commands i run that was fine so run the first two commands that you see here run these two now for the third one when you reach the third one this is what you're going to do i'm going to do cd space go back here right where we extracted it remember i just did it you can look on click here copy address says text cd space address all right we're now in the address now we can i'm an idiot copy address is text you want to copy the full address you see all of those files real real smart head and that's what i am whatever people make mistakes i'm not going to edit this out it's your fault if you mess up do something you're not supposed to just do something i don't know why we're uh i'm going to take it as all right dan you're so cool we already did that we're good all right we're in label img one of my favorite programs what are we gonna do alright we are going to want to go to where our screenshots are saved maybe this will work so open directory you want to cut very important click open direct not open this will open one image you do not want to annotate image by image very annoying open directory go in here paste directory boom we just we just did that yeah that's how cool i am screenshots select folder save changes directory you are going to want to go to the same place as where your images are stored so make it a whole lot less painful later on so what are we going to do now we're going to annotate our images i don't know i'm going to show you a few images that we're going to annotate and that's it okay so we have this running we're all good let's uh maybe you control and then scroll up to zoom in because images can be pretty small so um no nothing here i press d on my keyboard nothing there i press d on my keyboard d d i see this guy right here all right so if you know i click on him nothing happens hold down w ah now this appears and now drag boom we don't want that label this as person click ok do not create your own label also we can use this right here default label set it to person next what are we doing all right we save this ctrl s very important next it's going to go around labeling people control s next control s d so that for this process um it's just it's pretty relaxing honestly i am i do enjoy it one of my favorite parts is labeling these images i'm not even being sarcastic right now this is actually super fun to do remember to be hitting ctrl s because then you're gonna upload it to the program that i call roboflow well everyone calls it real buffalo you are gonna be without everything and all of your work will be for nothing actually i believe that if you annotate something and you click s oh yeah unsafe changes yes no so yeah got that error it's because you're not saving the images in there i see this is what i'm talking about not all your images have people which is why i told you to uh to get more images than you need it's not all have people i'm not gonna label that guy it's too confusing if there's like you know a guy he's right on the edge don't label him computer's gonna get hella confused dude so the reason we could label these images in roboflow which is the program we're going to go over next but it is super annoying to do that in that program i feel this is it you guys really don't need to see more of this so be back once i annotate 500 images oh my goodness guys oh i spent like an hour on that oh i'm sure you guys are tired of this too so luckily we are going to be able to sit back and relax for a couple minutes while we upload all of these photos onto the web just look at all these annotations by the way make sure that it was in yellow v5 i should have put it earlier in the video i completely forgot to tell you guys i had to redo everything that's why it took me so long i really hope you guys paid attention to that little arrow that i drew in very professionally all right so we are going to go open our web browser we're going to go to robo flow ai type that in roboflow uh we'll go sign up make your account so we're all doing with leaks my email absolutely catastrophe so as you've seen i've done this before right this isn't anything new what you want to do create an account right you should be at the screen right now maybe you'll get a few tutorial whatever who cares create new project all right project name what are we gonna name it we are going to name it i don't know how about let's do whatever your game name is and then however you know if you did it once put one if you did it twice because when cs go one right here do it twice that's what i like to do license you can keep this by default i'm going to put this mit which means anyone can use this and anyone can sell this um i'm nice like that just put person that's what we're detecting it doesn't really matter so yeah i labeled all of the person in these images who cares for example sells cars or pieces yeah great public project all right what do we do now i can tell you what we do now you have to upload it um don't so you can either do it through here or you can just go to screenshots right go to here and go back one folder by using that thing at the top you know click click drag it so processing files this is not uploading them to the web and make that very clear the process we're gonna have to wait for a while right now after we do this which is nice because i am totally exhausted all right returned annotation so they're fully within the frame perfect yeah so because we annotated this it may give you who cares continue all right finish uploading keep this as it is seventy percent trained twenty percent to test while training and then this is for the very very final one to tell us our accuracy continue 13 minutes so six seven take a couple minutes we get a sit back and relax so that's pretty nice see you guys five minutes tell you what you do from there all right guys uh so why we're all good here let's look at our data set training set yep we're all good here um generate new version that's what we want to click on once we go to a data set generate new version i i had a little mistake we all fixed it uh yeah i might actually edit out the mistake um i don't know so whatever train test split we're all good um this is not right i don't know why all the images that get uploaded but that that's not a problem for right now leave this as is argument leave this as is generate generate all right what do we do now i know we do now we go we google yolo v5 see see i've googled this before but we're not going to google that actually yolo v5 all right get up here we go google this we scroll down train custom data click here in this notebook google collab click here copy to drive now going to be your google drive click that do the captcha it's going to take a little bit let's go back here that's great so click the export button select a format scroll down yellow v5 part pi torch that's great um continue zipping files all right well that's sipping we'll go back to this one not this one because this is the one that doesn't belong to us so it's copy and drive this one this one belongs this to us looks a bit confusing that's because everyone's everything's already been done basically run this yeah i don't know what it does i just run it for fun really all right copy this you're not supposed to let anyone see this i don't know why i i don't i don't know what happens if someone does see it so guys please don't see it all right paste i'm gonna blur this out i guess i guess i'll have to kind of uh run so yeah that's it it looks like it's done i'm just gonna kind of put that folder over there up you guys can't see like i was saying again i i really don't know what that does but all right we're good we're at the training part images we've got our size like i was saying you can't change the size there's no there's no reason to do this but um batch 16 epochs this is not necessary we can just delete that one we don't you can train more that's just how many times it retrains don't have it at one because it's going to get like point zero zero six percent they'll have another one data keep this all good keep it in cash and loads it faster and time to train i should go all smoothly i'm going to tell you what's why and see here our utilization doesn't who cares honestly um so this is all being processed on google's and google is processing this for you it's pretty crazy how google just does that for free so right here you see this this is how accurate it is remember percentages are measured one equals a hundred percent so this is uh zero let's go up because zero from one percent zero point one zero point zero zero five percent that's what we're at right now you see as we're training more epochs this goes to point six now uh at this point where are we we are at three percent my 112 epochs the graph goes like if you go to actually it goes and then it plateaus and then it might slowly start spiking if it starts spiking that is bad that is bad if you have a very high confidence rate that is bad um like a very high like near 100 it's bad because that means you over trained it i mean unless you use thousands of images it means you over trained it and it's detecting just stuff just remembering all the boxes you don't want that to happen so it's training right now just training really fast guys compared to like other stuff and all we have to do is have the tab open this is insane actually this training speed so let's see where are we at right now six percent six to seven percent i regret not training oh no sorry not seven percent seventy percent we see this 70 60 70 70. 0.7 yeah you're so good 70 is pretty good you want to higher highest as possible well it's not like 98 unless you use thousands of images but like i said before i regret not running it for a little bit more epochs but epochs is how many times you rerun the code basically what's our best batch high 70's probably it doesn't seem that good i don't think that's that didn't seem that good around two minutes that took us results safe to run slash train slash epx we'll worry about this later right so you guys can run this for fun i don't actually know what it does it's not going to influence anything python detects which blah blah blah now if i believe this will show us our output all right guys sorry got it wrong um i got interrupted by someone yeah don't run don't run that run this who cares about this run this this will tell you 1993 yeah more that's the tip of my gun um person person person person person person person for training down 200 images remember glitch you know [Music] oh this is scary i don't know what this does this looks scary to me this is foreign so what do we do uh so we click these files yellow v5 runs uh detect we don't care we want train yeah esp won't wait this best file this can take a while click download okay while we wait we're also going to go here now what is here over here all right we're good download right here we have this discord invite it's going to stay up for a long time i hope octo yes that is how i imagine it's just not how the owner talks um so go to free downloads download whatever the current version is it's gonna bring you to mega should just be able to download it there shouldn't be a password if there's a password i don't know what to tell you if you're logged in log out that that helped me make sure this looks like that all right where are we downloads all right all right extracted so click on this install requirements this is probably going to take a bit longer for you guys because i already have everything installed we're good you might get an error you might get an error saying something to the effect of py presence bloody bloody block what you want to do if you get this you want to type in here like i did click here type in cmd that'll bring you into here what you want to do is you're going to install pip install whatever it's called if it's called like py presence i think p y whatever p y presence right whatever whatever if it says blah blah blah is missing on line 11 or something pip install whatever that thing that is missing all right if that happens might happen uh and well that will happen after you click this click run anyway all right so if you've got an error so that says something you know oh not found this chord p by persons do what i just told you take a while to load this is actually downloading it to cache just means i can easily access it all right great oh you see this this is nice right here this is if there's anything that detects it's going to put it here now to close this let's go here close that next what are we doing remember remember when we just downloaded that that file that best dot py file remember that i remember i remember um i also remember uh my brain i remember that too yeah i don't know where i was going with that so go into lib in the uh octo folder lib best.pt delete we don't need it that best.pt you downloaded from that google co-lab insert that yup that's right run it make sure there's no errors if you didn't change if you didn't add any new classes if you kept it as person it should run [Applause] now let's go back to our images let's go in csgo cs go what uh game play oh this is pinned i'm using desk pins so it always stays on top you don't have to do this i mean you can always just you know boom do it like that if you want to test it let's go images and people in these images come on detect them so mean of you i swear this actually doesn't work oh you see you detected him for like half us oh right there detects him trained out 200 images all the players have read out you can't blame me all right i'm sorry let's um actually oh we can get some of our native images that we have screenshots boom oh you see it detects him normal programs wouldn't be how i swear there's a feature where you can like arrow keys that's what it is boom detects him detect attack detect doesn't detect him guys i'm sorry it's not footprint all right this is which i changed it on 200 images it should be much better for you has 70 accuracy boom detected detected detected detecting detected detected guys come on it doesn't detect some you gotta remember this is like really crappy 200 images you're never supposed to train on 200 images all right uh surprise actually detected that guy mid-range that's where this thing does best so i hope you learned this but now now that you've done this and hopefully you've done it well if not you can always just you know go back to the normal file you know because you extracted it from the zip extractor from the sip again and you can just use it normally with a pre-trained model which is fine for most video games because you know people look like people what's your so actually the guy who made this if you're using the the config 1.2 yeah he forgot to go into that um version 1.2 he forgot to get to config and delete his config file so i'm actually going to want to run this again oh let's get rid of that old one all right and so see this asks us for game sensitivity yeah i mean just you know input but stuff honestly though zero point one one let's hope this works guys so the inbox i'm not sure how the hell you configure it i'm sorry i don't know i don't know i don't know how you do it press f1 to start i don't know how you do it what i do know but i do know for a fact is um very soon version two this guy its left uh right mouse button to flick yep and oh way out you know in in yes i also forgot to mention oh it's doing this freeze thing that's really not supposed to happen actually because that freezes your mouse breaks everything i hope it doesn't it probably won't happen for you guys it's like a new thing very annoying so basically you're gonna have to format it a lot but she's gonna come out with a trigger bot which just whenever you go over the enemy it shoots for you meaning that your accuracy is going to be improved you're not going to be wasting bullets which is really what everyone wants so guys thanks for coming to my ted talk or whatever it's been going on for way too long um wait for version two good good 200 fps no suffer from latency goodbye ladies and gentlemen i hope you learned something specifically how to do this what else would you learn from this except that i'm really bad at making this bye

Channel: DanTheEpicMan

Views: 21,038

Keywords:

Id: OyKwqTU1QRs

Length: 39min 25sec (2365 seconds)

Published: Sun May 08 2022