how to find robot in the real time with use camera?
camera on another robot that make movements.
Hi, your question is too general to provide support. You need to ask me something specific.
We have a problem
Need to find a robot in real time
What’s the best way to do this?
If the robot is not the same color
We used edgeimpulse, but it recognizes different parts of the robot, but we need the whole robot in a color image
So that our robot can always find another robot
Post example images?
Okay, so, are you looking down from the ceiling on the robot? What exactly do you mean by find it in real-time?
We don’t look from the ceiling. The camera is on the robot. We are looking for a robot while driving
A robot looks at another robot
I see, I would recomend usign the FOMO algorithm from Edge Impulse. It will work perfectly for this and give you a centroid of the other robot.
You just need to collect a dataset of images of the robot from the camera’s view. Like 200 images and then train a model. It will work well. You may want to get different images at different distances.
Yes, but during training, different parts of the robot are recognized, and not the entire robot itself
Used: FOMO MobileNet2 0.35
Number of training cycles = 20
Learning rate = 0.001
Validation set size = 20
Batch size = 32
Especially when our robot moves towards this robot"
“And there is another question.
Let’s say we have several cones (obstacles) of the same color. How do you think it is possible to organize their bypass? I was thinking about using the enumerate function to count the found cones and, depending on the screen resolution, somehow organize the movement”
Hi, yeah, you should take images of the robot at different angles, sizes, etc. The more the more robust the network will be.
As for dealing with color, use the find_blobs() method. It can do color tracking very well for you.
Is it possible to find a robot by key points if our robot with a camera moves around it? Can you tell me how to draw lines between key points and display their intersections on the screen?
Hey hi, isn’t it possible to implementing computer vision algorithms such as object detection or tracking.
Good afternoon, is it possible to create graphics in OpenMV? How to do this? I heard about QT-creater. How to organize their relationship? If anyone knows, I wouldn’t mind working together (just write your telegram)"
We have many drawing functions in which you can control the color of drawn objects. This allows you to create various things if you string operations together.
OpenMV IDE is based on QtCreator.