Frequently Asked Questions
This page lists common issues encountered during the installation and use of MIRA, along with their solutions.
Installation & Environment
Issue:
On Linux, MIRA fails to start with the error: Could not load the Qt platform plugin "xcb" in ".../cv2/qt/plugins".
This is caused by a conflict between the Qt libraries bundled with opencv-python and those used by MIRA’s interface.
Solution: You must force the use of the “headless” version of OpenCV, which does not include its own Qt binaries:
pip uninstall opencv-python
pip install opencv-python-headless --force-reinstall
Issue: MIRA runs on CPU even though a GPU is available.
Solution:
Ensure PyTorch is installed with CUDA support. Check your version with nvidia-smi and reinstall accordingly:
# Example for CUDA 11.8
pip install torch torchvision --index-url https://download.pytorch.org/whl/cu118
Performance & Usage
Issue: Crashes during high-resolution image processing.
Solution: * Reduce Image Size: Lower the input resolution in the settings. * Hardware: Ensure your GPU has at least 4GB of VRAM.
Issue: Model fails to detect objects or has too many false positives.
Solution:
* Thresholds: Adjust the Confidence and IoU parameters in the UI.
* Quality: Ensure microscopy images are well-focused and have consistent lighting.
—
Still having issues? If your problem is not listed here, please open an issue on the Issue Tracker.