DY112/LSMI-dataset

bug in preprocess_data.py

aamircs opened this issue · 3 comments

in preprocess data file all the obtained output image are black in color
it's not the white balanced image which we need for training the model

DY112 commented

Hi, @aamircs, thank you for your interest about our work.
The result image files of preprocessing code are not the general image, but the RAW tiff image.
So the range of pixel value will be 1023~16383, which is the bit depth of the camera which was used to capture the scene.

DY112 commented

I mean, all that black images are not complete black(0) , but they have 0-1023 or 0-16383 RAW values.
If you open that image files with cv2.imread function with the flag cv2.IMREAD_UNCHANGED, you can find that nonzero values are in there.
Since they are saved as lossless tiff format, general image viewer can not represent them properly to display.
If you want to see the scene content, please refer visualize.py and visualization process in 1_make_mixture_map.py.

However, the visualization process distorts the linear raw values and transforms them into sRGB format, so it is not suitable for use in experiments.

if VISUALIZE: