OpenCV Python Neural Network Autonomous RC Car
Вставка
- Опубліковано 11 січ 2025
- May 2015
Self driving RC car:
OpenCV neural network - Steering
Haar-cascade classifiers - Stop sign and traffic light detection
Ultrasonic sensor - Front collision avoidance
Raspberry Pi - Data streaming (video and sensor)
Arduino - RC car control
BGM:
[Hunter x Hunter 2011] Original Soundtrack 3
28 - Holding A Card File üm-ün
Apologize for the ad, at this moment, the bgm cannot be changed or removed.
Source Code:
github.com/ham...
zhengludwig.wo...
never wear a green outfit infront of this car
Tanvir Chowdhury lolololololoooloo
but it recognizes rectangular traffic light shape and recognizes green light later. Not all green things. It should be under that shape
@@science_mbg what is with programmers and jokes flying 3,000 miles above there heads
@@ThatGuyDownInThe coundt get it that he might be sarcastic my bad
@@science_mbg now I feel mean :/
Fake. There is a small person inside the car
lol i'm dying
😂😂
Oh! Are you in that car?
🤣😆😂🤣😆😂🤣😆😂
😁😁😁
Perhaps the only self driving Range Rover out there ;-)
I am studying Embedded System Development and I would find it awesome to go working for Land Rover. So maybe in the future there will be ;)
Lucid Gaming No please don't work for that awful car company. I'd rather work for a company which makes reliable cars.
TheGrayWolf81 Whats the problem you had with the brand?
They're just unreliable cars. My friend got a brand new one not too long ago and it's already having problems. Any other person I've known who has one has had problems with it (although some older models are nice and don't really have problems). I've never had much luck with English/British cars :/ But there are probably other people who love them and haven't had any problems yet. If I were to buy or do anything with a car, it'd be a Toyota or any other Japanese engineered car. They're the most damn reliable car I've ever owned and seen.
Why do you have a verification badge
Great Job!
Im your follower and like you
This game is amazing .
Check it out
ua-cam.com/video/MLcu3sBKhdE/v-deo.html.
@@cybercodes369 Why should I trust a Microsoft Spy?
@@benitokim1793 ok then keep sucking 😎
Wow awesome work! Seems simple to most people but there is a lot of logic going on behind the scenes! Love it!
People like you who perhaps do not have all the technical and financial support that could have, are the ones who truly make the future since with less they do much more, the truth my brother I congratulate you
congrats for this project dude. it's really brilliant and thx for sharing your work. u rock !!!
The impressive part is that this was made 5 years ago before all these easy to use libraries.
How did they managed to build it back then??
They mentioned Opencv
imagine making few of these cars and making mini city, then upscale this
It takes more time
The Car is a paid actor
Can confirm. I was the camera
*why don't you make an udemy course on this topic!!*
Because udemy is a platform that isnt to creators advantage, find a different platform
Sound pretty good
@@electronicsengineering7375 Thanks mate!
Hey, check this tutorial series to make your own autonomous mobile robot centaurirobotics.in/learn/
There is a github repo linked in the description. Just read the code
Wow first I was getting old videos in recommendation and now I am getting the same video that I watched years back giving me LOL and nostalgic feeling altogether.
This Guy: I finally did it.
Stackoverflow: [Leonardo DiCaprio DRINKING MEME]
PewPewZee LawL I'm not sure if you were joking or not or if you are a programmer but trust me, this was a joke lol and is probably correct
nice
I absolutely love the fact it's a red Range Rover. The most awesome vehicle out there and quite possibly its best color.
1:43 the fact that it knows that there is space in between the obstacle and itself is pretty great, I though it had to wait until the obstacle was out of the frame.
It's using Ultrasonic sensor to measure distance between obstacle and the car itself
i am computer vision engineer myself but this work that too 5 years past,,, this work of your astounds me
Nice work Wang it's good to see that you are sharing your knowledge with world.
centaurirobotics.in/learn/
This is the genius . Keep up the good work. This kind of videos motivates young programmers .
Hey Kiran, if you plan to make your own autonomous mobile robot you can check this out
centaurirobotics.in/learn/
please make an instructable
Shiva Kumar Read his paper.
Where can we read the paper
@@harshitsomani1717 It's the last link in the video description.
This is awesome! Please make a detailed video series on this project.
Whoah! That is impressive! In fact, incredible!
Looks great! Next step would be to get rid of the IR proximity sensor and detect obstacles via the camera alone.
Excellent work , image processing at its best . May I know how did you measure the distance between the car and Signal post ??
Measure the distance between the camera and the bounding box of the detected object.
Distance is one of the easiest things to do
@@wangzheng7526 Can you please make a tutorial on this
Good work! Way too good to see some of the common sensors act too smart!
Me: still
Print("hello world)
Output: Error EOF while scanning
Computer shutdown
Edit : oh need to edit this ... Guys I am trying to make a dumb joke and u guys are replying that" this is correct this is wrong"....... I know
print() can’t be capitalized. And I did get the joke.
I know guys what the error I am just kidding.
I am also a python programmer,
And about 'p' it's is becoz yt comments
Wait I got it........... you are passing a double float. Please convert to cell array and import fun functions. Can you believe I failed my programming course work?
def Print(text):
print(text)
Respect. I don't understand much about programming but i bet it was a lot of work. And it's awesome
Imagine having like 20 of these cars you could simulate road traffic with it
They would all interfere with each other as he uses an ultrasonic transducer on the front that can't handle that. And the car doesn't look for other traffic at the intersection. And so on ...
Arduino, sonar range finder, camera, opencv, pwm, lots of effort knowledge and intelligence. Nice
This is the task that many beginner python programmers feel like they're ready for after making a "Hello World!" program
Finally, technology and Hunter X Hunter combined.
:o I remember I saw this many years ago, and it was my inspiration to do an AI course and learn a lot. I never got to do this though, hope I can soon!
There is a MIT OCW course on AI. Course Number 6.034
It is FREE.
🙂
So who made this is a "man of culture as well".
P.s. The song in background is "Holding a Card File" from Hunter X Hunter Original Soundtrack.
Yeah, it's in the description!
Plot twist : A guy was controlling the car using remote
Great work!! Appreciating your efforts..
Hey Harikrishna, now you can make your own autonomous mobile robot. Check this out:
centaurirobotics.in/learn/
I am buying ruspberry.
Looks really cool.
Like.
Yes it stops at red and goes at green. The big question though is if it speeds up at yellow! On a serious notes, great job!
Incredible Peace Of Artwork There!!!
Great job brother I am truly amazed. I also thought of this kinda idea but didn't give it a shot.Great job man ,great job !!!!
Can you make step-by-step video tutorial for this? This is just amazing.
yes plz send step bystep this projcet
Sure , you just have to pay the price , you beggar!
ua-cam.com/video/GnnFy0uFe3E/v-deo.html
@@pperez1224 🙄🙌🤬
a good Autonomous car project. keep up the work!
stuart little was there in the car.
Great job! Excellent! One moment...
In real life if there is no stop line, you should stop near traffic light, in one line. But it's a small issue)) Good luck!
Damn! you are good, Google should hire you to collaborate on their Autonomous car project.
What's next? Autonomous drifting Nissan r34?
YurkerYT Haha, good observation! Maybe 86 as it's easier to drift... ?
Yea they would hire if he did it 30 years ago
Wang Zheng would you please share some tutorials. I am Python developer and I am interested in this field. please share some useful tutorials..
Bro once you learn raspberry pi learn opencv machine learning and neural networks you will able to create this and more than this don't learn to make only a make your base strong you will able to make thousands of projects
Except that good dont meddle in such basic things, and instead research more professional and complicated works
really nice work bro
i was thinking about it that someone will definitilly have done it and then i find this video
I am just so happy that there is no Windows.
Why did i Just start loving OpenCV and going to learn more about it?
No, no, no. An R34GT-R will never be an obstacle in front !
True
I like how this is build up. Using every single item you find in the household, ordenary cardboard, tape, womans hair bands... This is exactly how I would build it :D
Now you can recreate that GTA SA mission with the RC helicopter carrying barrels.
That was in vice city 😂
@Hot Dog Oh, one in VC was to carry bombs.
Wow a plend of neural and sensors it's great to watch
In Soviet Russia, the car drives you
LOL IDK Why Russian drivers are considered very bad
@@prateekpanwar646
I guess that joke could be switch to Communist China drivers now
In Communist China, the leadership drives you
пошел в жопу
@@matveynelson2489 АХхахаах Бум
@@timurbadretdinov9555 ?
So he used a Hunter x Hunter Theme to describe the epicness. Smart!
dayum computer science rules, I a mech eng student and seeing these just made me feel that I'm ignorant. btw I am truly interested in machine learning ai and robotics, hope to work with u in the future if we ever get the chance.
Congratulations for your awesome job ;)
"Mom, can we have tesla at home"
"We already have tesla at home"
Tesla at home:
Looks like it only stops for obstacles, because of the ultrasound distance sensor at the front not opencv.
Please make a step by step video on how to make this
Yes such tutorial will be a big step for the losers like us
step 1 learn python
step 2 learn aboun artificial neuronal networks ( the algo behind it )
step 3 code it , make mistakes , find them and solve problems
easy !
Bir Dost no such thing as l /w, just you are loserx or step by othersx, stupidx
Looking at the source code, i believe its a Haar Cascade
@@limitless1692 i know python ....and some of it frameworks...i want to do AI but i don't where begin
Amazing Interesting & Impressive. These types of video encourage me to code in python. Thanks! Keep share this types of ideas. Love. P.
Encourage me to also code in python
That's an incredibly great piece of work. I'm wondering what the car will behave when it sees a car with a red light or a "stop" sign on the back?
That nissan skyline gtr r34 just looks amazing
imagine elon musk copy his code for tesla car
Elon Musk has some slightly better hardware and software than a crappy Raspberry Pi and open source computer vision software.
Yt:lets reccomend this 5yrs later
It's 2018
Now that car is flying
It's 2020.. This car is in space now
Its 2020 june, the car has travelled to parallel universe and came back
This shows how good are autonomous cars
Actually this project is quite simple. It has been just elegantly displayed.
Okay then explain the control system architecture, how the CV and image processing works. Could be easy but I'd like to hear!
Epic, Thank you for releasing the source code!
Me:Watching a Remote controlled car yet it's called autonomous at 1am .
Tesla entered the chat .
Elon Musk is typing ......
Elon Musk: I need you at my Tesla Gigafactory in Storey County,Nevada
😂
Really amazing!! If it comes in real life it will better for our country .
background music is lit..🔥🔥
Should Hire this man for future cars
That’s how Tesla begin with the cars 🚙
came for the opencv, stayed for the tunes.
I found this project interesting. However, it seems that there is no motor control from the Arduino or Rasperry Pi. I guess you are transferring the wireless signal back and using the remote the control the car...
there is motor control for RPi ;)
NO You are not right
So why there is ultrasonic sensor in front of the car ?
These on a model train layout would be nuts
I read the description of your project on your wordpress and have a question if you don't mind: For your training dataset of input image and output direction input, did you hand label each of the images to have a certain output? For example, if the track is a slight left turn, did you take, assuming 10 fps, 30/40/50 images and hand label each to make a turn when you feel it is appropriate to make a turn? Thanks and great work!
I am curious to know about this too.. Is the training data hand labelled or you trained it by controlling the car via RC..
I just found out via his github and it confirmed my suspicions. He trains manually, saving a training image whenever he makes a keypress.
Nontheless still an impressive piece of engineering. Whatever you do in life there's room for improvement, he can improve the training.
Great.❤️ WHY IT IS RECOMMENDED 5 YEARS LATER ALSO.🔥 🔥
Wang Zheng
bro can you share the paper ?
the link you have given is not working
Nice work. I've made a vision based distance estimation using stereo camera. I really want to learn opencv and python when I do my research, but I learn slowly. It's look very cool. Hopefully I can learn the theory
Same
If you want to detect distance based on camera you need to use two camera to perform sensor fusion, just a single camera is not enough to measure a distance.
I'm curious to know why you chose mono vision instead of stereo vision?
He probably wanted to use that Raspberry Pi which only supports one camera at a time. Therefore, he used the ultrasonic sensor to detect depth instead.
Aterrebgnuoy you can use 2 camera using USB
Distance Measurement
Raspberry Pi can only support one pi camera module. Using two USB web
cameras will bring extra weight to the RC car and also seems
unpractical. Therefore, monocular vision method is chosen.
zhengludwig.wordpress.com/projects/self-driving-rc-car/
There was no need for stereo vision. Stereo vision of for depth sensing and wider FOV. The ultrasonic sensor does the depth sensing while the FOV is good enough. So no need for stereo vision.
FOV- Field of view
came for the engineering stayed for the song
what about keep left or right ? Like Australia is left drive, scan to left edge of road. ………. American is right drive.
This isn't funny.
Only one thing left to do, make it smaller, try to fit the gear under the body of the car, I would say its hard but looking at what you have done here I think you're up to the challenge.
Very impressive! 🤔
Very nice project. Congratulations!
That R34 though
Outstanding performance..dude we all will appreciate it if you make a step by step tutorial..
great , it needs speed limit.
Pretty cool, I was thinking about doing something like this back in 2015 and did some testing to see the feasibility of building autonomous racing robots with raspberry pi. The performance of python with openCV was too slow to process 30 frames per second based on simple benchmarks. I thought about off loading the image processing to a workstation, but decided to wait until raspberry pi got a faster processor. With raspberry pi 3 B+, I think it has just enough power to do it all on board without offloading the processing to a desktop or workstation.
+Wang Zheng Thank you Wang, Would you tell me you training code that you used to train you haar classifier? I am working on same project and this is my training code:
opencv_traincascade.exe -data hussein -vec sam.vec -bg neg1.txt -featureType LBP -numStages 6 -miniHitRate 0.999 -bt DAB -maxFalseAlarmRate 0.5 -numPos 900 -numNeg 2550 -featureType LBP -w 35 -h 62 -mode ALL -precalcValBufSize 1024 -precalcIdxBufSize 1024
and this is my makesamples code:
opencv_createsamples -info pose.txt -vec sam.vec -num 998 -w 35 -h 62
but my classifier is still wont work well.
last quesion:
How did you calculate minHitRate, numStages ,FalseAlarmRate, numPps and numNeg?
Thanks again
+Hussein Alaa
Here is the training code for stop sign (on mac):
opencv_traincascade -data data -vec samples.vec -bg negatives.txt -numStages 20 -minHitRate 0.999 -maxFalseAlarmRate 0.5 -numPos 800 -numNeg 400 -featureType HAAR -w 20 -h 20 -mode ALL -precalcValBufSize 5120 -precalcIdxBufSize 5120
You need more stages to improve the result, I chose 20 stages (in fact, at stage 16, the acceptanceRatio is 400 : 8.33416e-05, which means the result is good enough, a couple more stages won't help much).
Also, you might want to try HAAR featureType. LBP is good if you have high quality samples. The advantage of using LBP is not very significant in this stop sign case as it didn't take very long to train. Keeping your sample size small will reduce training time dramatically.
0.999 and 0.5 are the common values for minHitRate and maxFalseAlarmRate set in training. You can read some articles about adaboost to dig more.
Below shows the final training stage status of stop sign:
.
.
.
===== TRAINING 19-stage =====
Training until now has taken 0 days 0 hours 16 minutes 1 seconds.
Hussein Alaa I’m also working on a similar project. Can you send me the documentation for your project? I’m having problems with Numpy.
the algorithm explanation will be much better than music
PLZ SAHRE WITH US YOUR PROGRAMMING AND FROM A to Z, HOW TO MADE THAT CAR
that was really cool... awesome work
centaurirobotics.in/learn/
Can i use a normal webcam instead the raspberry one?
yes, but you should change the code in Streaming
have you tried the code??
Fricking awesome ! Hats off dude.
centaurirobotics.in/learn/
Plot twist : a person is controlling that car remotely
😂😂😂😂lol
Good lesson for those who are trying to pass their first driving test.
Most annoying music ever, though!
Dude wtf, I was just checking the comments to find out the name of the song and this is what I run into :D
its an hunter x hunter OST called holding a card file :)
Hello, your project is amazing, great job
Great job but prepare for the night ride because every car has red light on their tail
Tarzan in start up? Haha. Great jobs! Its actly 5 years ago.