Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Computer vision for drone landing pads #1268

Open
cbrxyz opened this issue Sep 4, 2024 · 5 comments
Open

Computer vision for drone landing pads #1268

cbrxyz opened this issue Sep 4, 2024 · 5 comments

Comments

@cbrxyz
Copy link
Member

cbrxyz commented Sep 4, 2024

What needs to change?

We will need to develop a computer vision model to detect the drone landing pads. @LesterBonilla found a camera that does automatic computer vision from some self-developed machine learning model, which we will try first. If that camera's built-in vision model does not perform at the level we desire, then we can develop our own ML vision model that will run directly on the drone's pi.

How would this task be tested?

  1. Ensure that the drone landing pad can be detected in a variety of environmental conditions, with the tins on it and not.
@cbrxyz cbrxyz changed the title Computer vision for drone landing pags Computer vision for drone landing pads Sep 4, 2024
@Josht8601 Josht8601 self-assigned this Sep 7, 2024
@cbrxyz
Copy link
Member Author

cbrxyz commented Sep 10, 2024

Hey Josh, I saw you were assigned to #1267 as well - can you just focus on this task for now?

@Josht8601
Copy link
Contributor

Josht8601 commented Sep 10, 2024 via email

@maxhv1 maxhv1 self-assigned this Oct 18, 2024
@maxhv1
Copy link

maxhv1 commented Oct 20, 2024

Status report: I attended the WAMV test on 10/16/24. I'll begin working on this issue soon. The plan for this, as suggested by @Keith-Khadar , is to create the landing pads in gazebo and develop a machine learning model based on that. I have the needed landing pad logos from RoboNation and the landing pad dimensions from the manual. I will make a set of images of the landing pad gazebo models, then label them in label studio, and then train a MIL Yolo model off of those images.

@maxhv1
Copy link

maxhv1 commented Oct 28, 2024

Status report:
image
I made the landing pads and cones in Blender, exported them for use in gazebo. I also labelled some images this week for small buoy detection for the animals task. I think I need to either make a new ros package and/or launch file for gazebo, since I think we want to run this gazebo world without the WAMV and try Gazebo with SITL for simulating the drone. I had tried running navigator_launch simulation.launch on a world file with the WAMV removed, it didn't go well. I tried installing Ardupilot earlier in the week also, but I mixed up which installation process to use (we need the legacy one for Ros1), haven't got that working yet. I think I messed up my gazebo doing that, I had to reinstall gazebo and now it looks like
image

@maxhv1
Copy link

maxhv1 commented Nov 4, 2024

Status report: I was unable to complete the gazebo world in time. I did help the mechanical team make a stencil for the Robonation objects for Task 8 (relevant to this issue), but I did not get to make a data set nor machine vision model off of it unfortunately. I tried to help figure out the params.py for path planner, but due to using a VM I was unable to run Gazebo and Nviz effectively.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

3 participants