anipose
anipose copied to clipboard
lambdaloop
Calibration Video Suggestion
Hi,
I was wondering if the board needs to be in videos in all of the cameras all the time. My setup has 4 cameras in 4 corners of a square area. So, it is not possible to have the boards projected on all the cameras for calibration.
I was also wondering if this is leading me to the error below:
D:\workspac_deepcut\local_DLC\FN 10-02-2020 - Short Clip-Chrissy-2020-10-08\videos\anipose_test\test1\calibration\calibrate1_cam2.mp4
99%|███████████████████████████▋| 3917/3960 [01:30<00:00, 356.40it/s][h264 @ 000002b22675ca80] Invalid NAL unit size (19244 > 789).
[h264 @ 000002b22675ca80] Error splitting the input into NAL units.
99%|████████████████████████████▊| 3934/3960 [01:30<00:00, 43.41it/s]
562 boards detected
D:\workspac_deepcut\local_DLC\FN 10-02-2020 - Short Clip-Chrissy-2020-10-08\videos\anipose_test\test1\calibration\calibrate1_cam5.mp4
100%|█████████████████████████████| 3960/3960 [01:00<00:00, 65.21it/s]
435 boards detected
defaultdict(<class 'int'>, {('2', '5'): 1, ('5', '2'): 1})
Traceback (most recent call last):
File "c:\users\rahmank\appdata\local\continuum\anaconda3\envs\anipose\lib\runpy.py", line 193, in _run_module_as_main
"main", mod_spec)
File "c:\users\rahmank\appdata\local\continuum\anaconda3\envs\anipose\lib\runpy.py", line 85, in run_code
exec(code, run_globals)
File "C:\Users\rahmank\AppData\Local\Continuum\anaconda3\envs\anipose\Scripts\anipose.exe_main.py", line 7, in
My config toml is given below:
project = 'anipose_test'
model_folder = '../../../'
nesting = 0 video_extension = 'mp4'
[manual_verification]
true / false
manually_verify = true
[calibration]
checkerboard / charuco / aruco
board_type = "charuco"
width and height of grid
board_size = [10, 7]
number of bits in the markers, if aruco/charuco
board_marker_bits = 4
number of markers in dictionary, if aruco/charuco
board_marker_dict_number = 50
length of marker side
board_marker_length = 18.75 # mm
If aruco, length of marker separation
board_marker_separation_length = 1 # mm
If charuco or checkerboard, square side length
board_square_side_length = 25 # mm
animal_calibration = false
fisheye = true
[labeling] scheme = [
]
[triangulation] triangulate = true cam_regex = '_cam([0-9])' cam_align = "1" ransac = true optim = true constraints = [
] scale_smooth = 25 scale_length = 10 scale_length_weak = 2 reproj_error_threshold = 3 score_threshold = 0.6 n_deriv_smooth = 2
Settings for a threshold filter
Removes data outside threshold (probably errors in tracking), and interpolates
[filter] enabled = true medfilt = 9 # length of median filter offset_threshold = 25 # offset from median filter to count as jump score_threshold = 0.8 # score below which to count as bad spline = true # interpolate using cubic spline instead of linear
Thank you for your help.
Did you solve this problem? I currently have the same problem. How to calibrate the cameras on the opposite side at the same time?
I think you need to have the board projecting on at least two cameras at the same time. That is how I solved mine.
Yeah, we've been rotating the board so it's visible in two cameras at the same time, like @KenR22 said. You can often tilt it forward and backwards at angle even when cameras are opposite to get them to see it.
It really is not ideal, and I would like to eventually add wand calibration to anipose, but i'm not sure when it will happen
@lambdaloop Would just like to +1 for your wish to add wand calibration. Although I know time is probably a limited resource, it would really make markerless tracking in human biomechanics studies much more feasible. And the value Anipose would add to a field that relies heavily on super expensive rigs would be tremendous. Anyways, thanks for all the great work on this!
