- Install XCB
sudo apt install xcb - tensorflow and ImageJ packages:
pip install -r requirements.txt- MaskRCNN-like repository
- When you found out the error like following below:
QObject::moveToThread: Current thread (0x5578e7485030) is not the object's thread (0x5579015f8b40).
Cannot move to target thread (0x5578e7485030)
qt.qpa.plugin: Could not load the Qt platform plugin "xcb" in "/home/min20120907/anaconda3/envs/cell/lib/python3.10/site-packages/cv2/qt/plugins" even though it was found.
This application failed to start because no Qt platform plugin could be initialized. Reinstalling the application may fix this problem.
Please remove the qt library from opencv-python by:
rm -rf $PYTHON_PATH/lib/python3.10/site-packages/cv2/qt/
If you run this within 4K or higher resolution, you can type this environment variables:
QT_FONT_DPI=96 QT_SCALE_FACTOR=2
If Segmentation fault occurs, please also type this environment variable:
LD_PRELOAD="/usr/lib-linux-gnu/libtcmalloc_minimal.so.4"
If the cause is memory fragmentation maybe the environment variable 'TF_GPU_ALLOCATOR=cuda_malloc_async' will improve the situation.
The usage of new "coco_convert_gpt.py":
python coco_convert_gpt.py --append_mode --coco_path '/SSD-1TB-GEN4/dataset-png/train/FIB-SEM of a dividing cell at 3.9 min after anaphase' --zips_path '/SSD-1TB-GEN4/dataset-png/train/FIB-SEM of a dividing cell at 3.9 min after anaphase/single-2000-2200-RoiSet.zip' --mode single --txt .tif
The usage of evaluation:
python eval_model.py --dataset /SSD-1TB-GEN4/dataset-png/ --workdir . --weight_path logs/cell20230527T1017/mask_rcnn_cell_0024.h5