southwestroboticsprogramming / tagtracker Goto Github PK

Could it be possible to add rotation to the solver and update the network table?

The current install script uses apt by default, which is against best practice, the dependencies should be instead listed in the install section of either the wiki or the readme. It should also not assume that teams want to use anaconda, and should allow the use of requirements.txt. The install script could actually be full removed if other people agree with this.

The apriltag library that the code is using is no longer being updated while pupil-apriltags (or similar) is. Changing over should make the code more reliable and faster.

Add a template Systemd service that teams can use.

Can you add a detailed explanation for the transformation matrix of the camera?

There is a chance that a future game will have apriltags posted on a moving game element. We should be prepared for this by adding support for apriltags without a defined location.

pip install apriltag requires cmake to be installed, so on Ubuntu, we had to sudo apt install cmake.

requirements.txt and the documentation are missing two pip dependencies: imutils and flask_opencv_streamer

The JSON being put out for distortion coefficients is not formatted correctly (missing commas):

  "dist": [[ 2.96051672e+01 -1.15858321e+04 -1.79874268e+00 -1.79700977e+00
  4.78984685e+05]]

Looks like some of the sample files have been corrected, but incorrectly; they were changed from a (1, 5) matrix to a (5, 0) matrix (caused by dropping the outer brackets).

Logitech Webcam file:

{
    "fx": 676.6192195641298,
    "fy": 676.8359339562655,
    "cx": 385.1137834870396,
    "cy": 201.81402152233636,
    "dist": [ 0.01632932, -0.36390723, -0.01638719,  0.02577886,  0.93133364]
  }

should be "dist": [[ 0.01632932, -0.36390723, -0.01638719, 0.02577886, 0.93133364]]

Suggest fixing calibrate_camera as:

        print('  "dist": {}'.format(dcoeffs.tolist()))

Like the title says, the apriltags detector is only trying to detect the 36h11 tag family when the 16h5 family should be detected instead.

This allows the robots to calculate their position on the 2023 game field.

I'm looking through the different tags that you guys set in your environment and I'm wondering why you guys have the tag size at 8 inches and not 6 as FRC changed it to that. I'm also wondering about the rotations of the tags. Isn't the frame of the apriltags z into the tag, x right of the tag?