Hi,
First, let me thank you for this code and the paper. I think this is a great problem and I like your solution.
However, I found that there are issues with the ramp_detection function, even on the sample data: it sometimes detects small peaks that are far away from the actual ramp.
Let us inspect how the start of the flash is detected currently:
# wrong
start = np.flatnonzero(profile[:max_pos] > ramp_detection_thresh)[0]
The problem is that it will return any small peak above ramp_detection_thresh from the beginning of the image to the max of the profile (which is inside the flash).
Instead, it should detect the first value below ramp_detection_thresh from the max of the profile to the beginning of the image:
# ok
start = max_pos - np.flatnonzero(profile[max_pos::-1] <= ramp_detection_thresh)[0]
Similarly, for the end of the flash, the current code is:
# wrong
end = max_pos + np.flatnonzero(profile[max_pos:] > ramp_detection_thresh)[-1]
but should be:
# ok
end = max_pos + np.flatnonzero(profile[max_pos:] <= ramp_detection_thresh)[0]
Do you want to do the change yourself, or should I publish a PR?
Hi,
First, let me thank you for this code and the paper. I think this is a great problem and I like your solution.
However, I found that there are issues with the
ramp_detectionfunction, even on the sample data: it sometimes detects small peaks that are far away from the actual ramp.Let us inspect how the start of the flash is detected currently:
The problem is that it will return any small peak above
ramp_detection_threshfrom the beginning of the image to the max of the profile (which is inside the flash).Instead, it should detect the first value below
ramp_detection_threshfrom the max of the profile to the beginning of the image:Similarly, for the end of the flash, the current code is:
but should be:
Do you want to do the change yourself, or should I publish a PR?