CV2D converts optical images of hBN flake to height profiles. It uses simple 2-layer dense neural network based on the tensotflow platform.
Tensorflow 2.3.0 is required for this project. It requires python 3.5, 3.6, 3.7, or 3.8. Simple way to ensure correct version is installed is to create separate environment
conda env create --file environment.yml
conda activate cv2d
or equivalently
conda create -y -n cv2d python=3.8.15 tensorflow==2.3.0 pillow==9.3.0 opencv matplotlib==3.5.3
conda activate cv2d
You may also use pip
conda create -n cv2d python=3.8.15
pip install tensorflow==2.3.0 pillow==9.3.0 opencv-python matplotlib==3.5.3
conda activate cv2d
If you want to use this environment in jupyter-lab, create kernel assosiated with this python environment
pip install --user ipykernel
python -m ipykernel install --user --name=cv2d
After you've installed tensorflow module try running the script with test image (provide no arguments)
python cv2d.py
You need to provide flake image and background image. For example,
python cv2d.py input/img1.jpeg input/bkg1.jpeg
You may also specify background color in RGB format using square brackets:
python cv2d.py input/img1.jpeg "[175, 115, 155]"
Or just flake color and backgroung color (valid padding is used for this case)
python cv2d.py "[89, 151, 213]" "[175, 115, 155]"
Pay attention to quotes around the argument and use integers between 0 and 255 to specify the color.
Result is saved to output<#> folder.
Keep in mind, that non-square images are automatically rescaled.
You may open grayscale image with Gwyddion and provide z-scaling.