Is it possible in Python to read an image and get each pixels color (12 bits depth) value between 0-1 (or 0-255 in float) without doing the conversion to 0-255 -> 0-1 range. This way I won't get precision loss and get the exact color in HDR format.
It look like cv2 and PIL read pixels only in integer value.
Here is a good repo of Paul Rohde that I would like to use in example for image : https://github.com/codelogic/wide-gamut-tests
Updated Answer
You can open a proper HDR file like this to get it in full float32 scaling:
Original Answer
I still don't understand your question. The image you have shared is 8-bits per sample, or RGB888. It is neither 12-bit nor HDR.
You can see that with
exiftool, 6 lines from the bottom:If you load the image with OpenCV, you can see the text that you say you cannot see: