General toolbox for 4D Particle Tracking Velocimetry. Allows you to track in space and time hundreds or a few thousand of lagrangian tracers in a flow.
This completes my last PR. Most importantly, I thing the code is now correct, at least
With both Python versions, the same number of matches are found out of the same number
of candidates: "6754 matched found (out of 135271 candidates)".
However, I didn't use the Matlab function to check that. @jsalort you could try the new Python version and check whether the results seem correct.
In terms of performance, the Python version is now only 10% slower than the C++ version (compared to 20 times slower before these PR). Some improvements of Pythran could help in the near future.
There are still some very simple tasks to be done on this code, in particular using argparse and h5py (or h5netcdf) to save data in a better format.
This completes my last PR. Most importantly, I thing the code is now correct, at least
However, I didn't use the Matlab function to check that. @jsalort you could try the new Python version and check whether the results seem correct.
In terms of performance, the Python version is now only 10% slower than the C++ version (compared to 20 times slower before these PR). Some improvements of Pythran could help in the near future.
There are still some very simple tasks to be done on this code, in particular using
argparseandh5py(orh5netcdf) to save data in a better format.