Hybparc Aruco

Hands on ECG-placement Training software for med students.

-> To have the software save the annotated pictures, create a 'results' directory. (Care, per pass 1 pic meaning rn per analysis 10-15 pictures, uncompressed!)

General requirements

We recommend running this software in its own environment, e.g. conda

• OpenCV
• Numpy
• PyQt6
• ArucoRoi (Must be manually put in /lib/)¹
• qt-material (for styling only, but has to be manually removed from the code if unwanted)

1: Correct the import error in detector.py by changing import services to from . import services.

Linux specific adjustments

Video capture id:
Windows and MacOS use id 0 for the first camera, 1 for the second and so on. On Linux a USB camera has multiple dev/videoX interfaces, e.g. video0 to video3 for the first and video4 to video7 for the second camera. We recommend using the lowest interface per camera, i.e. 0 and 4. Using the Win/Mac 0 and 1 indices will result in an attempt to grab from the same camera, crashing the software.

Video capture resolution and format:
On Linux OpenCV (/the V4L2 driver) defaults to something, which is likely the lowest resolution and a wrong/limiting capture format. This can be fixed by setting the cv capture properties by hand, e.g. if our camera supports 3840x2160 via MJPG, the following should be applied to all your VideoCapture objects:

stream.set(cv.CV_CAP_PROP_FOURCC, cv.VideoWriter.fourcc('M', 'J', 'P', 'G'))
        stream.set(cv.CV_CAP_PROP_FRAME_WIDTH, 3840)
        stream.set(cv.CV_CAP_PROP_FRAME_HEIGHT, 2160)

To check available formats, resolutions and capture indices run v4l2-ctl -d /dev/videoX --list-formats-ext

OpenCV and Python version:
On MacOS and Windows PyQt6 supports Python 3.12. On Linux however, only Python 3.9 is currently^{Dec '24} supported. You will have to downgrade your environment to Python 3.9 or you will be unable to install PyQt6.
Additionally, potential issues may arise from which OpenCV package you install. On MacOS the standard opencv-python package works well. On Linux not everything might be supported. In case of crashes we recommend opencv-contrib-python.