Lane detection and object detection with OpenCV & TensorFlow
Вставка
- Опубліковано 31 бер 2018
- From webcam feed, the program will identify the lane marking using OpenCV.
The motion tracking is via Lucas-Kanade optical flow.
docs.opencv.org/3.4.1/d7/d8b/...
The object detection is using TensorFlow Object Detection API.
github.com/tensorflow/models/...
Music: www.bensound.com - Наука та технологія
คนไทยสุดยอดครับ เป็นไปได้ ขอศึกษาด้วยครับพอดีกำลังศึกษาเรื่องนี้อยู่ :)
This is so cool! Well done! :D
That, sir... Was amazing!
Very awesome update on your previous video!! Seems like you have been working hard at it, the interface looks very polished. Also appreciate your use of GH and open source, I'm sure the beggars will be working their way in blindly asking for the code. Looking forward to the next one! Cheers.
Oh wow. That's a lot of pride with as if you just made it to the moon.
- Approaching truck on left Lane
- "Turn left"
WebDev feel like its just tellin you to stay centered in the lane the whole time, not really to change lanes, at least thats what i get from the crosshair in the middle of the screen
@@lrdyeah To be fair I didn't notice that thing in there lol
0:37 see the reflection on hood...
Where's the issue?
@@amp08021 motion tracking in reflection on hood!!
Weird! ig!
@@imkronos_me I don't think this should cause any problems though
@@amp08021 i think the video is edited!
Of course... the algorithm can detect the reflections, but if that would be real, it looks different... Logic
I'm getting the idea that I need to build this into my windshield.
Only programmers and engineers understand this .
It was amazing and interesting
Probably yeah. My girlfriend does NOT see anything exciting here.
It looks so trivial to most people just before they know how to drive a car with their hands.
Why would a normal person watch computer vision video
Mathematicians developed all the fundamentals for this. Engineers only make use of the library someone wrote. I'm saying this as an engineer who loves to actually understand how the algorithms work.
@@est9949 who cares?
Congratulations for your works
I want to know what Fusion With Steering Advise feature has been smoking
I really wanted to see him change lanes
Could you tell me how much delay you have? That's from the time you begin to take the picture to the time the program identifies all the objects?
This is simply WOW
its soo cool bro! Hat's off!!!
Is there any way to build a dashcam out of this?...I have been looking for a solution the whole day long and all I got is a major headache
That might sound weird, but watching this gave me the same feeling as watching some Superhero Movie where the power is released/revealed (or Dragonball Power ups😂)
Awesome examples, just beautiful! Thanks ❤
Awesome Work Man...Keep Going.....
If you make some tutorials on this then it will be very useful for most of the Computer Vision enthusiast and beginners
Open CV is open source. This guy didn't create it.
Hey @@jamesbarker6373 !!!
My interest is to know how he did the object detection.
That's why I mentioned that tutorial for object detection like he did.
@@rajharry1378 Google OpenCV. They have tutorials there.
@@rajharry1378 docs.opencv.org/3.4/d7/d8b/tutorial_py_lucas_kanade.html
@@@jamesbarker6373
Thanks....
Fusion with steering advise is in track motion?
I already do that with my brain that’s why I’m Subaru’s finest rally driver
Niubility awesome
😒oh..god
It would have been nice if the the video was not sped up. What is the actual framerate you are getting at full resolution?
1-2 fps :P
ผลงานเยี่ยมมากครับ ขอชื่นชม แต่ถ้ามาชนบท เส้นขาวตีมั่วไปหมด 555 เดาว่าคง error marking
But why driving in the second lane when not passing?
Goodjob mate! Btw, can you tell me which processor or pc that you used for processing thing? Thank you)
How do you get it to do something like turn on a servo or turn on a light if it detects something like a car?
เจ๋งมากคับ
Great ❤️
Let's see what it does in construction zones where the old lane markers veer hard right into a barricade.
It would rely on object detection to know that action would be stupid.
How can I find same input video?
Can I find documentation somewhere? I would like to lern this myself, and this is the best solution I could find so far.
u can learning machine learning and deep learning he is opencv which is computer vision library for object car and lane detection u can learn it do learn machine learning and deep learning
does it work on rain?
I guest , it is depending on the training data set
For the objective detection, either YOLO algorithm or Faster region proposal CNN algorithm can handle that well
Indeed. Tensorflow use Faster R-CNN, see the github repo
@@mauriciocortazar9604 Why not YOLO instead ? Is there any advantage to use Faster RCNN that I dont know ?
@@user-wd8wx5md5z Althought YOLO is faster than Faster R-CNN, YOLO fails detecting smaller objects. Another reason is because YOLO is more recent than the former.
@@mauriciocortazar9604 YOLO V3 is now better at detecting objects of 3 different scales, including smaller objects due to some more downscaling layers added to it. This makes it slightly slower though. But you can't have both.
correct me if im wrong;
Using hough transform to detect the lines;
using tensorflow api for object detection;
and then what?
please advice me on how to implement the steering advice method.
all i can break from the video is : you are using the 2 lines to get the mid point between the 2 lines. and how does that work?
ua-cam.com/video/821ZnewFLcY/v-deo.html
Watch this and view description for codes, I hope you will get clear understanding about this concept.
is there any way that you can run this through a mobile phone? I mean using your mobile phone's processing power. its a bit sci-fi but I want to make a helmet with an fpv screen. having camera's outside. day camera, night camera, infrared/thermal camera. oriented front, rear and sides. not showing all camera's at once. but you can switch between them wherever you want.
it doesn't need to identify every single thing. doesn't have to recognize the road. just vehicles and pedestrian.
Is there any other method except averaging out left and right lane separately ?
Inform lanes of each other and admit a maximum curvature, a minimum distance between lane separation lines and inform lane detection of likely lane shape, width and curvature from a GPS and a map. Many things you can do. Stuff gets messy when you have sun glints, reflections, wet asphalt, differing colors of concrete (bridge, roads repaired) or other lines such as tire marks. Honestly, pure OpenCV for this is not enough. You'd need a neural network too.
Awesome
it's cool ,bro
how to do this in google colab,I want to use colab to do just like this
Can u suggest the object detection course
how do you use the code? maybe you can make a tutorial
Please send the code of lane detection
Sir I need tagged vehicle type dataset can you share a source?
I wander what the motion tracking is used for and how you accomplished it. That's about my research, please tell me. Thx
Google motion tracking. As simple as that
So cool.
Can you please help me to get the source code for learn it.
is this the real speed of the video when detecting some lane? or anything that you present in this video??
In the object detection case, processing one frame takes ~6 secs on CPU and ~0.8 on GPU,. The lane detection is fairly fast in terms of processing time. I don't know about the motion tracking though. Probably what he did is record the video and then post-process it...
@@sebastianliendo6579 would mind check this code?? I am having a problem it consumes so much memory around 2gb and above i don't know why,
private void timer1_Tick(object sender, EventArgs e)
{
using (Image nextFrame = captureIP.QueryFrame().ToImage < Bgr,byte>())
{
if (nextFrame != null)
{
// There is only one channel (gray scale), hence the zero index
// var faces = nextFrame.DetectHaarCascade(haar)[0];
Image grayframe = nextFrame.Convert();
var faces = haar.DetectMultiScale(grayframe, 1.3, 4, /*new Size(nextFrame.Width / 8, nextFrame.Height / 8)*/new Size(420,300));
foreach (var face in faces)
{
nextFrame.Draw(face, new Bgr(0, double.MaxValue, 0), 3);
}
pictureBox1.Image = nextFrame.ToBitmap();
}
}
}
private void Form1_Load(object sender, EventArgs e)
{
string facePath = Path.GetFullPath(@"../../data/haarcascade_frontalface_default.xml");
haar = new CascadeClassifier(facePath);
captureIP = new VideoCapture("rtsp ip cam link");
}
yea but how long to prcoess that video.....
❤️
I want the tutorial..anyone can help me?
Are you able to measure distances too?
What is your set up?
Whats the music ?
You guys need to add a time decayed persistence of objects in your process
how in the world did you got verified with 28 subcribers
@@Bl4ck309 Go get a life. Damned if I know!
สวัสดีค่า เห็น Test ถนนในไทย อยากจะขอคำแนะนำหน่อยได้ไหมคะ ^^
Aimbot activated
ติดตามครับ อยากได้ความรู้เพิ่มเติม
Please Can you make complete tutorial on this how to make it
And explanations
Can we have the full Surce Code
i need good course to make a good project like that
How it's work?
Kindly share a video with live coding please
Ya, do that on a foggy day...
It seems that some of the predictions are dangerously wrong...
Dear sir, How fusion of opencv and tensorflow works in real time with a webcam. Can you tell me the process? I am not asking for code but I wanna know how fusion works. or you have used the tensorflow output as the input for opencv for fusion. Kindly reply and guide.
Mayank Raj fusion?
He probably means: how to use a TensorFlow model in OpenCV.
Oh? If that's the case it's a simple case of acquiring your data through openCV and using that as training data for your tensor flow model. Think of openCV as a glorified camera that takes snaps
For example, the lane detection done is done using Lucas-Kanade method(irrelevant) and the cars are detected using tensorflow.
@Ivar van Wooning This is guider video ua-cam.com/video/mm8GWRRn9L0/v-deo.html
can you make tutorial how you achieved these huge achievemenst ?
Google Open CV
I think this features are not in Tesla car
Nice work
fps ?
Source code please
i did similar thing on Carla simulator
source code?
What is motion tracking?
It tracks objects movements.
Posted on 1 april, coincidence?
are that project need database
Cool, change lanes, thas really cool, lol
Hi I am working on OD project using Tensorflow OD api but getting an inference time of 0.22 second for single frame can you tell me how did you detect with this speed you are showing remainder I am on CPU
if you are using a cpu instead of a gpu then that's the reason.
So people see turn left and assume it means 90 degrees? It could mean turn left .01 degrees.
turn right :))))))))
sir where from we learn about it
Where I can find tutorial about tensorflow and opencv together?
That really depends on what you want to do with it, any more specific details?
hey thanks , i want to analyse video data with tensorflow
@@mrlive221 then what do you need opencv for?
@@LeahLundqvist ,i want to use tensorflow with webcame like for gesture recognition
@@LeahLundqvist , do you know howto use both ?
code source ? svp
This is definitely not in real time and rendered in after the footage was recorded... Still not bad but nowhere near "Tesla level".
Tesla is still not bad but nowhere near space shuttle level.
why not? you only need extra capacity for video coder
Did you use exactly the code from the example?
can you share your code?
The direction was wrong
can you share me your code? this is it's a good project
с вертолета с машины обнаружили устройствам передали и т.д. это мы не умеем.
Like technology opencv
car's ap is too low
code
Hello it was amazing could you help me to make this project
thaiii
*--*
Judgment day is coming soon guys lolz
1:19 bruh
Too sentimental..
ช่องของคนไทยป่าวครับเห็นชื่อเหมือนคนไทยเลย
Just Google Comma.ai
Great.., can u share source code
But it's really really easy. Why do you want the code?
did u got the output?
@@priyanandamuri226 yeah
can u give me ur contact number?if u dont mind...because iam doing this as my final yearB.Tech project..i want some help :) plz rply me fast
@@allmightqs1679 no.
Check this course for an in-depth explanation of OpenCV from beginner to advanced: ua-cam.com/video/gLDJEY49M38/v-deo.html
чушь какая-то.
I don't think it's accurate enough. Your lane tracking method still needs to be optimized.
Python
*d