Hi,
I want to use OPENMV Cam H7 in a box that can be teached.
So I put the object 1 ( a black tube ) in front of the BOX and press a button so the Camera save this object on it’s memory.
Object 1 : https://www.sandleford.com.au/content/images/thumbs/0004239_rail-modular-matte-black-tube-25mm-900mm_480.jpeg
Then, I put the object 2 ( a black thermal sheath ) in front of the BOX and press a button so the Camera save this object on it’s memory.
Object 2: https://encrypted-tbn0.gstatic.com/images?q=tbn%3AANd9GcTG6t3YUC_ydS05nUsu4XWLjyV9RQjQeXGOj9dK_iyCBjG9lnbi&usqp=CAU
And then I put the object 1 into the object 2 and then teach the camera the right position between them by pressing the same button.
The camera must recognize the right position between them and If I put the two objects in the right position again The camera fixe the two tubes by two servo-motors.
I want to know if this is possible with OpenMV Cam H7 or not ?
Cordially.
Yeah, it’s possible, I know of some customers who have done this. However, you need to be advanced enough to modify the firmware and write your own Machine Learning code in C.
Thank you sir:)
Just another question please. It cannot be programmed with python ? in OPENMV IDE ?
Most parts, you need a custom ML algorithm for this however… which requires performance.
Thank you.
Could you please give me where to learn that ? I 'm a beginner in this field.
And, is the basic open mv h7 cam is enough or I need a global shutter shield and buy MT9V034C for that because I need precision or I must buy open mv H7 cam plus ?
Hi, I would suggest you learn how to code the algorithm on a computer first before thinking about how to get it onto the OpenMV Cam.
Generally… coding a custom ML algorithm is pretty hard. You generally need to be a pro to do this.
Given your project, maybe checkout PyImage Search and learn more about computer vision.
Thank you sir.
I will need the custom ML algorithme to make the camera recognize the black tube and the black thermal sheath ? Or to teach it ?
And, is the basic open mv h7 cam is enough or I need a global shutter shield and buy MT9V034C for that because I need precision or I must buy open mv H7 cam plus ?
Thank you
I don’t know how to.
You’re asking for a feature that is possible on the camera but only if you really know what you are doing. That said, we don’t have this type of feature right out of the box. Given that, you may want to use a different product that has this feature by default.
The only thing we have onboard the camera that’s similar to what you want to do is FAST Keypoint Matching and ORB Keypoint matching. These algorithms aren’t robust compared to CNN models available nowdays however.
We are working on an automated ML pipeline that will be out in the coming months but we are not there yet right now.
Jeez, that was annoying read. I’m sorry you all have to deal with that.
