Hi all,

It seems cdtw doesn't support running DTW with two sequences of feature vectors (e.g. MFCC). I get the following error:

Traceback (most recent call last):
  File "./s2f.py", line 161, in <module>
    d = U[b] - V
  File "./s2f.py", line 88, in __sub__
    compute_path = False)).get_dist()
  File "/my/path/to/python/python2.7/site-packages/cdtw/pydtw.py", line 12, in __init__
    cydtw.__init__(self, ref, query, settings)
  File "cdtw/cydtw.pyx", line 251, in cdtw.cydtw.cydtw.__init__ (cdtw/cydtw.c:4379)
  File "cdtw/cydtw.pyx", line 279, in cdtw.cydtw.cydtw._dtw (cdtw/cydtw.c:4989)
  File "/my/path/to/python/python2.7/site-packages/numpy/core/shape_base.py", line 278, in hstack
    return _nx.concatenate(arrs, 0)
ValueError: all the input arrays must have same number of dimensions

The code excerpt generating this error:

        return pydtw.dtw(this.stFeatures.T, that.stFeatures.T, pydtw.Settings(step = 'p0sym',     #Sakoe-Chiba symmetric step with slope constraint p = 0
                                window = 'palival', #type of the window
                                param = 2.0,        #window parameter
                                norm = False,       #normalization
                                dist='euclid',
                                compute_path = False)).get_dist()

this.stFeatures.T and that.stFeatures.T are numpy arrays and have shape (38, 13) and (65, 13) respectively.

Is this feature not supported or did I do something wrong?

Any help is appreciated, thank you!

I get this error with the speedtest and here.

(Pdb) dtw([1,2,3],[2,3,4],Settings(dist='manhattan', step = 'dp2', window = 'nowindow', compute_path = True))
*** TypeError: an integer is required

I wonder if you are interested in creating a package for pypi? That would be great, as it would greatly simplify installation (pip install cdtw) and usage.

examples.ipynb :
import pydtw

compared to
Readme.md :
from cdtw import pydtw

Readme.md:
The hyperlink of
"well known step patterns dp1, dp2, dp3"
does not link to the intended PDF.

Hi,

I wonder if you could add a license? See also:
https://help.github.com/articles/open-source-licensing/
http://choosealicense.com/

MIT would be great ;)

C:\Users\p\Downloads>python test.py
Traceback (most recent call last):
File "test.py", line 280, in
run(s,t)
File "test.py", line 63, in run
s.step.set_type('dp2')
AttributeError: Step instance has no attribute 'set_type'

C:\Users\p\Downloads>python test.py
Traceback (most recent call last):
File "test.py", line 280, in
run(s,t)
File "test.py", line 67, in run
d = dtw(ref,query,s)
File "C:\Users\ap\Downloads\cdtw\src\pydtw.py", line 11, in init
cydtw.init(self, ref, query, settings)
File "cydtw.pyx", line 250, in cydtw.cydtw.init (cydtw.c:4337)
self._dtw(ref, query, settings)
File "cydtw.pyx", line 259, in cydtw.cydtw._dtw (cydtw.c:4588)
c_dtw_settings.dist_type = settings.dist.get_cur_type_code()
TypeError: an integer is required

This is resolved with:
python setup.py build_ext --inplace --compiler=msvc

TEST PASSED
C:\Users\p\Downloads>python test.py
Ran 20 tests. 0 failures.

Hi,
I have an error when I import pydtw after the installation of cdtw:

from cdtw import pydtw

Return:
ImportError: cannot import name 'cydtw' from 'cdtw.cydtw' (unknown location)

Can anyone help me?

Hi!

I don't understand the following behavior (code adapted from examples nb):

import numpy as np
import pydtw
t = np.arange(0,3*np.pi,0.2)
r = np.sin(t)
q = np.cos(1.2*t+1) + np.sin(t)
print(pydtw.dtw(r,q,pydtw.Settings(dist = 'manhattan', step  = 'dp2', window = 'nowindow', compute_path = True)).get_dist())
print(pydtw.dtw(r,q,pydtw.Settings(dist = 'manhattan', step  = 'dp2', window = 'nowindow', compute_path = False)).get_dist())

Output:

12.37014033416319
11.829838028295049

Why do I get different distance when setting compute_path to True or False? Shouldn't the distances be the same regardless (maybe I'm missing something...)?

Thanks!

C:\p\SonicEvents>lib\python\python.exe speedtest.py
Traceback (most recent call last):
File "speedtest.py", line 16, in
s.step.set_type('dp2')
AttributeError: Step instance has no attribute 'set_type'