For anyone wondering: the parameter epsilon in the method approxPolyDP represents the maximum distance between the shape and the approximation. If epsilon = 0 then the approximation will be really precise. But the bigger epsilon gets, the further you allow your approximation to be less rigorous, and an epsilon that's too big will simply not detect shapes properly. Trying the code as it is in the video on some random shapes image I found on google, I get that 0.01*cv2.arcLength(contour, closed=True) is actually an approximation that's "too good" (as in it'll try and draw more than 3 edges on a triangle). I put epsilon = 3 and it seems to do the job fine. Depending on what you're coding and how precisely you want to detect, play around with the epsilon value.
@@theearthish The aspect ratio was used because of the noise. Under normal situation, the width and height of a square is the same, in that situation the aspect ratio will be 1.
Can anybody tell me why it is multiplied with "0.01" at 5:40 ,Line no: 10. I was liking his way of explaining things but totally disappointed now. Not explaining basic only -_- 👎
Did u listen to the whole thing? The closer your epsilon value is to 0, the more precise is your approximation of the curve and so your shape will be detected more precisely. Learn to listen.
@@shubhambagdare6600 Did u listen to the whole thing? Where it is mentioned "The closer your epsilon value is to 0, the more precise is your approximation"?? Btw thanks for clearing my doubt.
what if the object is a vehicle or plant? we can definitely identify them but how can we name it? like we used 3 sides for triangle and so on. What can be the logic to name car and say cactus?
No matter what I try, I get the error - "too many values to unpack (expected 2)" on the countours definition line. I am running directly from the repo and with the downloaded image file noted in comments.
It is because , the findContours will give 2 outputs. May be you are giving like (contours = cv2.findcontours(" ") ). You have to give like( contours, _ = cv2. ........).
For anyone wondering: the parameter epsilon in the method approxPolyDP represents the maximum distance between the shape and the approximation. If epsilon = 0 then the approximation will be really precise. But the bigger epsilon gets, the further you allow your approximation to be less rigorous, and an epsilon that's too big will simply not detect shapes properly.
Trying the code as it is in the video on some random shapes image I found on google, I get that 0.01*cv2.arcLength(contour, closed=True) is actually an approximation that's "too good" (as in it'll try and draw more than 3 edges on a triangle). I put epsilon = 3 and it seems to do the job fine. Depending on what you're coding and how precisely you want to detect, play around with the epsilon value.
You're awesome. Your Github Gist works
We can also use
width == height instead of using aspect ratio, it will work as same as the ratio and it is simple to understand. (If w == h: "square")
I think aspect ratio is used so that it takes care of any error in the boundingRect method
@@theearthish The aspect ratio was used because of the noise. Under normal situation, the width and height of a square is the same, in that situation the aspect ratio will be 1.
Can't wait the next one!
Thanks for this great video!
where do I find the shapes.jpg file? I could not find it in OpenCV github
I did it using paint. It took me 2 min. All the shapes are there.
Thanks a lot, wonderful!
great. Thank you for this video.It helped me a lot.
Does it works
Can anybody tell me why it is multiplied with "0.01" at 5:40 ,Line no: 10.
I was liking his way of explaining things but totally disappointed now.
Not explaining basic only -_- 👎
Did u listen to the whole thing? The closer your epsilon value is to 0, the more precise is your approximation of the curve and so your shape will be detected more precisely. Learn to listen.
@@shubhambagdare6600
Did u listen to the whole thing?
Where it is mentioned "The closer your epsilon value is to 0, the more precise is your approximation"??
Btw thanks for clearing my doubt.
@@adityaanand5686 You are fool someone is helping you and you are giving reply to him in negative way instead of Say Thanks
Thanks for sharing!
If with the above problem, I want to identify parallelogram, what should I do? Looking forward to your support. Thank!
How can we detect something which is combination of two or more geometries, like an arrow?
Thanks for video
Will this python program work in real time geometric shapes detection.??
Can you help me with the "trackbar" to find the threshold values?
Hello. Can you please tell how can we detect the same in live video.
I'm getting error in line 5 .. Should I need to save the 'shapes.jpg' to a specific path?
you need to enter the full path with image name and extension. Example: "E:/saurabh/image/shape.jpg"
@@reealtech_xr thanks bro ❤
How to get there x,y coordinates values
where could I download the shapes.jpg for testing? Thanks a lot
Hi guys I was just wondering if I could use this code for nuts shape classification and detection?
any help is appreciated.
Thanks a lot
thanks bro......
hi thanks for the tutorial
can you tell us how we can do it in real time ? not just for an image ,and thank you
what if the object is a vehicle or plant? we can definitely identify them but how can we name it? like we used 3 sides for triangle and so on. What can be the logic to name car and say cactus?
maybe car would fall in the category of trapezoid.
That falls under object recognition
do you have a live version of this which can detect from webcam , please
cap = cv2.VideoCapture(0)
can i use this for movement shapes
sir what is the aspect ratio for rectangle because i want do for trapozoids also...!
Can u know aspect ratio of rhombus?
You saved me woahhhhh from pepe's fury
How can i get there x,y coordinates values?
Did you ever solve this?
any luck in getting shapes x, y coordinates?
No matter what I try, I get the error - "too many values to unpack (expected 2)" on the countours definition line. I am running directly from the repo and with the downloaded image file noted in comments.
i think you run it in python 2 not 3 or you didnt upgrade the version of opencv
It is because , the findContours will give 2 outputs. May be you are giving like (contours = cv2.findcontours(" ") ). You have to give like( contours, _ = cv2. ........).
Try with "_, contours, _ = cv2.findContours(thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)"
Too late but it can probably help someone in the future.
I can't find the shapes image. Help?
неужели нельзя приложить файл с этой е.бучей картинкой?
how to detect the color of a simple shape in opencv ? could you please give me any references?
You here for eyrc?
@Dandavati You can change the image to HSV, and then use value of Hue to detect the color.
What does ravel does
As per my knowledge , ravel will give the starting point of contour draw in img pixel. So x will give x- coordinate
Y- gives y coordinate.
@@Its-Not-Funny do this code works
How to find colour of the shapes? Can anyone help
Are you from Nirikshuk bot ?
@@RiteshNEVERUNIFORM yeah
@@RiteshNEVERUNIFORM I sort out this one but I am struggling in task _b
Did this code help u to solve nirishuk task 1
Instead of displaying the shape name once, it shows multiple times. Anyone faced the same issue?
I faced the same issue. Did you find a solution for it ?
I had that too, the reason for mine was i didn't use elif for the other if statements after the first one
Im also facing this problema, and im using elif statements though
@@gauninha12 help??how?
Reduce the threshold. It is working perfectly after reducing. I changed mine from 240 to 225.
there are way too many ads in between.
You're only reading through the code. Total waste of time.
Have you watched the previous videos? Most of the functions are explained there.