Add calibration calculators #2609
Conversation
added PlainExtractor based on numpy and scipy functions
restructured the stats containers
…ctapipe into stats_extractor
Remove StarVarianceExtractor since is functionality is featured in the existing Extractors
allow overlapping extraction sequences
renaming to chunk(s) and chunk_size and _shift added test for chunk_shift and boundary case
| ) | ||
| # Check if at least one chunk is faulty | ||
| if np.all(valid_chunks): | ||
| raise ValueError( |
There was a problem hiding this comment.
Is this really an error? Or should it just return as a no-op?
There was a problem hiding this comment.
The check is needed because if all chunks are valid, an empty list would be vstack later on, which results an error without a proper message. I would prefer to break here and put a check for the validity of chunks outside this function, to avoid having multiple returns in one function. I find it really hard to follow the logic of a function with multiple returns (maybe just a personal taste). If you insists I can modify it to return a no-op. However, if everything went well in the first_pass, the second_pass should not be called in my opinion.
|
After this, we are still missing a And that will finally replace the two classes in |
Correct, we have prototype tools ready in
Yes, we would have to make sure that the interface to |
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
polish doctrings for trait help
Analysis Details0 Issues
Coverage and Duplications
Project ID: cta-observatory_ctapipe_AY52EYhuvuGcMFidNyUs |
This PR contains the
StatisticsCalculator, which aggregates statistics, detects outliers, handles faulty data chunks. TheStatisticsCalculatorholds two functions to conduct two different passes over the data with and without overlapping chunks. The first pass is conducted with non-overlapping, while overlapping chunks can be set by thechunk_shiftparameter in the second pass. The second pass over the data is only conducted in regions of trouble with a high percentage of faulty pixels exceeding the thresholdfaulty_pixels_threshold.related to issue #2542