openptv / postptv Goto Github PK

In scene.py it.izip is used in line 34 and 457. This should be changed from "it.izip()" to just "zip()" to be compatible with python 3.x

`(test) MacBook-Pro:postptv alexliberzon$ pytest tests/
============================= test session starts ==============================
platform darwin -- Python 3.7.7, pytest-5.4.1, py-1.8.1, pluggy-0.13.1
rootdir: /Users/alexliberzon/Downloads/flowtracks/postptv
collected 25 items

tests/test_analysis.py .. [ 8%]
tests/test_interp.py ........F. [ 48%]
tests/test_io.py .... [ 64%]
tests/test_scene.py ....... [ 92%]
tests/test_sequence.py .. [100%]

=================================== FAILURES ===================================
____________________________ TestJacobian.test_inv _____________________________

self = <test_interp.TestJacobian testMethod=test_inv>

def test_inv(self):
    pos = np.array([[0., 0., 0.]])
    tracer_pos = np.array([
        [0.001, 0, 0],
        [-0.001, 0, 0],
        [0,  0.001, 0],
        [0, -0.001, 0],
        [0, 0,  0.001],
        [0, 0, -0.001]
    ])
    # Basically we interpolate something based on the average position
    # change, because it's easy for me to visualize.
    interp_data = tracer_pos*2

    interp = interpolation.interpolant('inv', 6, 3)
    interp.set_scene(tracer_pos, pos, interp_data)

    local = interp.interpolate()
    np.testing.assert_array_equal(local, np.zeros((1, 3)))

    jac = interp.eulerian_jacobian()

  self.assertTrue(np.all(jac[:, [0, 1, 2], [0, 1, 2]] != 0))

E IndexError: index 1 is out of bounds for axis 1 with size 1

tests/test_interp.py:170: IndexError
=============================== warnings summary ===============================
tests/test_io.py::TestPtvis::test_trajectories_hdf
tests/test_io.py::TestPtvis::test_trim_hdf
/Users/alexliberzon/opt/miniconda3/envs/test/lib/python3.7/site-packages/flowtracks/io.py:710: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
dtype = np.dtype(fields)

-- Docs: https://docs.pytest.org/en/latest/warnings.html
=========================== short test summary info ============================
FAILED tests/test_interp.py::TestJacobian::test_inv - IndexError: index 1 is ...
=================== 1 failed, 24 passed, 2 warnings in 4.53s ===================
`

AttributeError                            Traceback (most recent call last)
Cell In[3], line 12
      7 # Here we start to use the repeated-interpolation machinery,
      8 # By informing the interpolant of the current frame data,
      9 # and then querying it about that data without having to repeat it.
     10 interp.set_scene(frame.tracers.pos(), frame.particles.pos(), 
     11     frame.tracers.velocity())
---> 12 neighb_base = interp.which_neighbours()
     14 # Check that we have a particle with the desired number of tracers.
     15 candidates = neighb_base.sum(axis=1) >= 10

AttributeError: 'InverseDistanceWeighter' object has no attribute 'which_neighbours'

when running repeated_interpolation example

Hello Yosef and Alex,
Thank you for the valuable package you provided.
I am using flowtracks for Lagrangian studies and I have two questions I hope you can help me with those:

1. I get the length of each trajectory by using following line:
   traj.pos()[-1,0]-traj.pos()[0,0]
   the maximum length I get is 0.16 but when I plot traj positions I see the maximum length of the trajectories is equal to the length of the observation volume (0.3 m).
   my guess is that the plot function just join small trajectories and make a longer one or so. If yes, based on what criteria it does that? Please correct me if I am wrong.
   2)is there any class for Lyapunov exponent calculation in flowtracks?
   Thank you for all the help,

This could be a necessarily trigger for us to do everything in postpv. I think as a first thing a gallery of Jupyter notebooks with some demos could be a good option.

I think we need to change the way we use indexing in trajectory.py

from

self.__dict__[attr][selector]

to

self.__dict__[attr][np.array(selector).astype(np.int)]

for instance particle_pairs fail for the call of the type:

trajid = [0,1,2]
particle_pairs(original,shifted,trajids=np.array(trajid),time_points=np.zeros(3))

with the error

/Users/alex/miniconda3/envs/postptv/lib/python2.7/site-packages/flowtracks/pairs.pyc in particle_pairs(primary_trajects, secondary_trajects, trajids, time_points)
     52         prim_in_frame_ids = np.unique(trajids[coord_locator])
     53         prim_in_frame = [t for t in prim_traj if t.trajid() in prim_in_frame_ids]
---> 54         prim_parts = take_snapshot(prim_in_frame, frame_num, schema)
     55 
     56         sec_in_frame_ixs = trajectories_in_frame(secondary_trajects, frame_num,

/Users/alex/miniconda3/envs/postptv/lib/python2.7/site-packages/flowtracks/trajectory.py in take_snapshot(trajects, frame, schema)
    305         first_frame = traj.time()[0]
    306         for prop in copy_keys:
--> 307             kwds[prop][trix] = traj.__dict__[prop](frame - first_frame)
    308         kwds['trajid'][trix] = traj.trajid()
    309 

/Users/alex/miniconda3/envs/postptv/lib/python2.7/site-packages/flowtracks/trajectory.py in getter(self, selector)
     66                 return self.__dict__[attr]
     67             else:
---> 68                 return self.__dict__[attr][selector]
     69 
     70         def setter(self, new_val, selector=None):

IndexError: only integers, slices (`:`), ellipsis (`...`), numpy.newaxis (`None`) and integer or boolean arrays are valid indices

probably also in other places, e.g.

Hi @yosefm, just a question, why when we look at pairs in some frame, we request for trajectory to be active also in the next frame?

I guess we have urgently move to python 3
@yosefm could you please help us with the direction?

i've found that flowtracks.io.trajectories_table has been using older PyTables convention
http://www.pytables.org/MIGRATING_TO_3.x.html

e.g. alexlib@5cf008c

One of the common issues that can be solved through post-processing is the linking of the broken trajectories. There are publications by Beat Luethi (called glueing) and Haitao Xu (multi-dimensional tracking) that dealt with this problem for 3D-PTV specifically and there is an old Matlab code that I've wrote some time ago

Matlab linking code

The idea is rather simple - for each two trajectories that are possibly broken pieces of the same trajectory project the edges (last for the earlier and first for the later) towards each other and see if they are within the acceptable tolerance (of distance, velocity and/or acceleration and time difference). If they are, connect, replace them by a single peace, make it smooth again, repeat the search.

The example notebook in /examples https://github.com/OpenPTV/postptv/blob/master/examples/hdf5_scene_analysis.ipynb

apparently uses inverse weighting. but I get NotImplementedError. Looking into the file interpolation.py I do see that it's not implemented.

https://github.com/OpenPTV/postptv/blob/master/flowtracks/interpolation.py#L703

Shall we fix the example or complete the inverse weighting implementation?