AI Destroys a Mobile Game in Real-Time with OpenCV (Object Detection) 

It's annoying when people don't read up on something at least a little before making stupid comments

I really appreciate this comment. Honestly though if it gets more people into it I can take a little heat. No brain no pain haha... Thanks for watching and commenting.

Thank you!

I actually thought this was just a well made program too, i knew that an Ai is something that needs training such as machine learning and involving generations, so i went to the comment to find an explanation since i'm just a begginner in the programming world, so could you tell me what makes this an Ai?

@@FireDotExe Of course. Although note I"m not claiming to be a leading expert in the field so take this with a grain of salt. AI is simply the simulation of human intelligence in machines. So in this program... playing the game using simulated mouse inputs! We did this using computer vision and rules. Machine learning on the other hand you are using more of a black-box approach. So you're not telling the computer how to do the job it's learning how to do it by itself.

I have never been more disappointed with myself. I did however have to move on and code again.

😞😞😞😞

Top 10 saddest anime moments 😢

train the ai to get 69420 points and pause in all games

Changed my mind don't turn off internet keep watching my videos. Thanks for view and comment 😉.

Yea it like stores ads for a bit

Apps like this download 2-3 videos to play if your Internet is off

Also some say that you need to the internet in order to play but its only for the ads

Guy you may be, but them cheeks gonna feel what it means to get balls to the walls.

Great minds. Thanks for watching and commenting.

Why have you played zigzag for years?

yes i too always play games and wonder how ai would do in it !!!!!!!!!!!!!!!!!!!!!!!!

Honestly comment means a lot thanks so much.

Thanks that really means a lot to me! Thank you for watching and commenting.

Darn AI.

The first R in "RETRY" also had a circle, the second one however only a line. I think you'd include a minimum distance two objects must have to be considered different objects. You can also see early on how it didn't detect an edge up to the point where it touched the ball but actually had a gap there.

I honestly really appreciate every view and sub. Thanks for commenting and subscribing.

Comment really means a lot to me. Thanks for taking the time to watch and leave me a message.

Thank my man! Really appreciate the sub.

I'm on it. Thanks for watching and commenting.

Thanks! I really appreciate the comment and view.

Thanks for sub Emil means a lot.

This is a VERY smart comment. A few people have mentioned the idea and it crossed my mind very late in making this video. I decided I didn't want to invest the time to try it but I think it would work.

I think that on top of this you could detect the intersection point between the left and right lines (if you just made the lines go off to infinity before drawing), and with a static distance from that intersection point determine the ending point of the line. This would in theory fix the opencv "interference" behavior and draw more consistent lines. It'd also make the damn thing perform perfectly

Yeah, I was surprised when he used a hough-transform, as something similar to Sobel with different kernels (or straight up, what was its name, canny?) should have sufficed quite well (and not resulted in infinite lines). Would be edges and not lines, but it seemed like he used the lines for edge-detection anyway (just painted "hazards" [with binary directions] for the algorithm to trigger direction changes with).

hey thanks for watching I appreciate it.

Honestly this crossed my mind very late in this project. I think it's a good idea! Thanks for watching / commenting.

Probably the easiest way is to simply remove all vertical lines, all lines that are connected to the underside of the vertical lines. After cleaning the lines, you can also remove the gaps in the edges, I think this will also help. As the next steps, you can also check that there is a full path from the current point to the end of the screen. And clean everything between the two edges ( artifacts from gems for example ).

Awesome I will! Thanks for watching and commenting.

I do feel bad for those people part of the reason why I picked stupid not so popular games.

@@ClarityCoders I'd say the people who are able to bot those games applied their own skill to the game.

Hola, hablas español? Me dirías si pudiste hacer funcionar este proyecto? Tengo que hacerlo para un juego casi igual

Thanks JKL that really means a lot. I appreciate you taking the time to comment.

Thank you RU-vid overlords. Thanks for watching and commenting my man!

@@ClarityCoders no problem ^^

Listening to Rick rn

I'm glad you like it. Should be plenty more coming soon.

That's a really good idea someone else mentioned it as well. Towards the end of the project I thought about finding all lines and only keep ones at the angle of the track. It was late and I was over it at that point so I never tried.

Dang I wish I had thought of that before posting it was on my old laptop. Next time I will post that as well.

I love playing around with it. Thanks for the comment.

Thanks! Never to late to get involved start small and work your way up!

something like sethbling's marIO would work for that i think

Yup probably need to go deep learning but it could be done. I'll put it on my list.

Instead of wait maybe a month to see it happen why do you try making it in about a year! Much faster

Thanks that means a lot. I got more cool stuff coming out soon!

@@ClarityCoders im exited for that

Probably boat loads of tweaking and huge training sets. But it is definetly possible and the programs became better at doing these text based captchas than humans so we switched to Googles weird meta data and image recogintion ones. And im 100% convinded that Google uses those "select all X" pictues as a pretense to build image recognition sets. They can do it just fine without us having to cllick anything

Thanks dood. Keep coding!

uwu

@@volttheprotogen1893 uwu

Thanks! That's a spot on analogy haha.... Overall though I appreciate the comments and views even the ones complaining about my haircut 😂😂.

I'd love to try that game! haha we could use the sound as input right?

@@ClarityCoders Yeah of course! Search "Dancing Line" on Google play. This got me thinking about the connection between the visual representation of music and the AI's ability to "see" things.

Thanks!

Thanks for watching and commenting!

@@ClarityCoders sure, y not?

Thank you for watching I really appreciate it.

I'm all about the balls.

Thanks Zack means a lot.

It's a fun challenge for sure. Thanks for watching and commenting.

@@ClarityCoders I always comment on videos I want to see more of. Got to train that algorithm!

You might break terms of service and get banned but nothing about it is illegal. Note I'm not a lawyer and this is not legal advice. 😉

flappy bird?

Whew thought I lost you there glad you got it back! haha

Thank you!

Thanks! I appreciate the view and comment.

That game’s still going I think, god it had such bangers, the design of each stage only got better as you went along! That Easter egg in the maze was heartwarming too

@@explosivemonkey4596 ok but still can't play it because i using android

@@ductr2hn I pray for your loss.

@@explosivemonkey4596 they flooded it woth ads at some point it got unplayable for me

Thanks RU-vid has been fun for me doing stupid projects. I'll stick around if a few people keep watching! Thanks for comment/view.

Thanks! Really glad you enjoyed it. Lots more coming soon.

Probably what's the game called?

Mean a lot! Thanks for commenting Legend!

Those were the days. Cheat engine baby.

You're certainly better than me.

Fast and Run - Nico Staf. Thanks for watching and commenting!

Thanks! I've been wanting to do a line by line one for a project like this I'll put it in the works.

Probably 8 hours spread over a few weeks? Basically it was really simple to create the rules and automate it to play repeatedly but dialing in parameters is tough. So I would change some things then let it run for 100 games then change some things.

@@ClarityCoders The next step: use a genetic algorithm to look at the best parameters hahahahah

@@Diego-Garcia Then we would enter machine learning territory!

Thank you!

*visual confusion*

It did! It slept so well that night all tired from it's big day.

Yeah I've had this comment a few times and you guys are correct. I think that would be the easiest way to improve the project.

@@ClarityCoders honestly though thank you for responding I had been wondering. Great videos too! Edit: now subscribed

You could try and show us 😉

I'll check it out and put it on my list!

OpenCV

@@ClarityCoders Thanks! I was thinking it OpenCV was only a library or something.

Cooking the competition.

the answer is yes

Scrap II is pogging so hard. 10/10 besties!

It's not there yet but it's getting easier for non programmers to do as well. It will be something that is an issue in gaming in 2 - 5 years. Not that it isn't already but it will be so wide spread.

LoL no. A good anticheat can detect whether you're cheating or not based of your aim. A human aim is bit shaky makes errors, a robotic aim is very precise and can be pretty easily detected

@@ShivamJha00 You would have to mimic pro-level accuracy. Basically intentionally making it shaky and a bit random.

Amaizing job btw

I'll put it on my list of games! Thanks for watching / commenting.

Hope you're right 😉. If not glad people are enjoying it! Thanks for watching and commenting.

Fast and Run - Nico Staf

You are a stud. I literally never improved I must not have the attention span or reflexes in my old age.

@@ClarityCoders Its super repetitive in a bad way, like wait until the turn and just tap once or you lose and when multiple turns appear you can easily lose track of how many times you have to tap. I just like everything that can be challenging and try it until I win it or at least know how to play it well. The code was more interesting as always compared to me just mastering this game hahaha.

or you could just play a game that doesnt require an ai to play it for you? you want to "play" a game without actually playing it because the game sucks that bad...

@@godw1ll99 why even click on this video if you have that mentality? lol

@@KingofEvil this video is achieving an entirely different goal for a completely different purpose. Its almost 0% comparable to my comment. In this video there was never any desire to play the game the intention was strictly to make an ai. The commenter here has the desire to play a game but doesnt actually wnt to play it. One example is to work to create an ai the other is to use an ai to cheat in a game for him out of laziness. I could try to clarify further but im already being pedantic.

@godw1ll99 where exactly do you see the comment saying he wants to play the game? All he said was that he wanted to automate certain actions in popular games which are in the „p2w“ genre which might just be a experiment to see how a ai compares to a p2w player for all you know. I do not see a reason for you to lecture someone when you do not even know their intentions correctly. Absolutely pathetic.

@@KingofEvil my comment was addressing what he said specifically meanwhile you fabricate hypotheticals in order to argue with me and you call me pathetic? he made clear as day what he was wanting to achieve with the first half of his comment. he wants to automate the core functions of a game leaving nothing left but the false satisfaction of having achieved something in the game which makes the entire endeavor meaningless. its no different than people that bot a game like rune scape. the insanity of having the desire to achieve something in a game but not wanting to actually play the game is whats pathetic. "to perform basic actions such as "gathering resources", 'auto attack/defend/whatever' because of the mind-numbing actions you have to perform and the fact that it's literally P2W. It's hard to socialise with people ingame with that sort of system." is the most blatant example that he fully intends on "playing" the game with a side of "automation" to do the grinding for him.

I'm amazed people watch my videos let alone comment on them. I truly appreciate every one of them even the ones making fun of my unibrow😂. Thanks for the comment and the view!

Yeah buddy!

Can you click on anything? Try something other than the game maybe.

@@ClarityCoders I actually figured out there is a setting in the emulator that wasn’t letting pyautogui click

An AI that is written in C could do it, but Python is too slow to process that much data in rt

Great question. Even with processing full image data in some games I was able to achieve better than human reaction times. Although as someone mentioned there are faster language choices out there if speed is an issue.

Could maybe do a follow up on it. Thanks for commenting and watching.

By definition it's artifical intelligence though it's not machine learning. I would agree with your point though it's not learning anything.

AI does not require any form of learning.

1. AI 2. Machine Learning (a type of AI) 3. Deep Learning (a type of machine learning and AI)

anything that makes decisions to solve a problem is intelligence, and if it is created by humans we call it artificial your phone has several ai systems, even route planning is AI the edges of the definition of ai is fuzzy, but for example the game itself is not ai, although if there are enemies that try to hurt you, or even if the track generator uses tiles you die on more often later in the level and less often earlier, that is ai

@@Maric18 as you said the term AI is fuzzy and what we perceive as AI also changes with time. It's similar to technology. Some time ago a pair of scissors could be considered a technology but today we have a much different idea of what technology is.

I get what your saying but actually Zig Zag was one of the hardest way harder than Fall Guys. You need to have such a quick script and react to the changing colors of the environment. Thanks for watching/commenting I really appreciate it.