When running multicamera calibration: "ValueError: non-broadcastable output operand with shape (1,3) doesn't match the broadcast shape (3,3)"

[joshkodiaklabs]

Appears similar to #374

Figured I'd create a new issue and include all of my data. I've experienced this with both 2 and 8 camera setups. And have tried using single image sets as well as multiple image sets. Tarball with raw images, bag, and target yaml attached. Also included the commands I am running.

kalibr_issue.tar.gz

Potentially confounding bits:

1. I get "Initialization of focal length failed. Provide manual initialization". I have been entering 2 (assuming it's in mm). I believe the actual focal length is 2.75mm, but I haven't found it myself via calibration

Setup

$ git show master
commit 1f60227442d25e36365ef5f72cd80b9666d73467 (HEAD -> master, origin/master, origin/HEAD)
Merge: 82597ff 0478139
Author: Patrick Geneva <goldbattle@users.noreply.github.com>
Date:   Fri Mar 8 12:56:02 2024 -0500

    Merge pull request #673 from AndyZe/write_svg
    
    Add an AprilTag option for svg
$ docker build -t kalibr -f Dockerfile_ros1_20_04

Bag creation

rosrun kalibr kalibr_bagcreater --folder /data/images/images/april6/stereo/ --output-bag /data/images/images/april6/stereo//out.bag

Calibration

rosrun kalibr kalibr_calibrate_cameras --bag /data/images/images/april6/stereo/out.bag --topics /cam0/image_raw /cam1/image_raw --models pinhole-radtan pinhole-radtan --target /git_repo/april_6x6.yaml --bag-freq 10.0
importing libraries
Initializing cam0:
	Camera model:	  pinhole-radtan
	Dataset:          /data/images/images/april6/stereo/out.bag
	Topic:            /cam0/image_raw
[ WARN] [1750113795.873714]: BagImageDatasetReader: truncated 0 / 1 images (frequency)
	Number of images: 1
Extracting calibration target corners
  Extracted corners for 1 images (of 1 images)                              
Initialization of focal length failed. Provide manual initialization: 
2
Initializing focal length to 2
	Projection initialized to: [nan nan nan nan]
	Distortion initialized to: [nan nan nan nan]
Initializing cam1:
	Camera model:	  pinhole-radtan
	Dataset:          /data/images/images/april6/stereo/out.bag
	Topic:            /cam1/image_raw
[ WARN] [1750113799.932455]: BagImageDatasetReader: truncated 0 / 1 images (frequency)
	Number of images: 1
Extracting calibration target corners
  Extracted corners for 1 images (of 1 images)                              
Initialization of focal length failed. Provide manual initialization: 
2
Initializing focal length to 2
	Projection initialized to: [nan nan nan nan]
	Distortion initialized to: [nan nan nan nan]
initializing initial guesses
	 initializing camera pair (0,1)...  
Traceback (most recent call last):
  File "/catkin_ws/devel/.private/kalibr/lib/kalibr/kalibr_calibrate_cameras", line 15, in <module>
    exec(compile(fh.read(), python_script, 'exec'), context)
  File "/catkin_ws/src/kalibr/aslam_offline_calibration/kalibr/python/kalibr_calibrate_cameras", line 465, in <module>
    main()
  File "/catkin_ws/src/kalibr/aslam_offline_calibration/kalibr/python/kalibr_calibrate_cameras", line 219, in main
    baseline_guesses = graph.getInitialGuesses(cameraList)
  File "/catkin_ws/src/kalibr/aslam_offline_calibration/kalibr/python/kalibr_camera_calibration/MulticamGraph.py", line 176, in getInitialGuesses
    success, baseline_HL = kcc.stereoCalibrate(cameras[camL_nr], 
  File "/catkin_ws/src/kalibr/aslam_offline_calibration/kalibr/python/kalibr_camera_calibration/CameraIntializers.py", line 32, in stereoCalibrate
    r_median = np.median(np.asmatrix(r), axis=0).flatten().T
  File "<__array_function__ internals>", line 5, in median
  File "/usr/lib/python3/dist-packages/numpy/lib/function_base.py", line 3501, in median
    r, k = _ureduce(a, func=_median, axis=axis, out=out,
  File "/usr/lib/python3/dist-packages/numpy/lib/function_base.py", line 3410, in _ureduce
    r = func(a, **kwargs)
  File "/usr/lib/python3/dist-packages/numpy/lib/function_base.py", line 3556, in _median
    return np.lib.utils._median_nancheck(part, rout, axis, out)
  File "/usr/lib/python3/dist-packages/numpy/lib/utils.py", line 1177, in _median_nancheck
    result[n] = np.nan
ValueError: non-broadcastable output operand with shape (1,3) doesn't match the broadcast shape (3,3)

[joshkodiaklabs]

Following #635 if I initalize to 200 it converges and I don't get this error.