cadop / pycgm Goto Github PK

View Code? Open in Web Editor NEW

23.0 4.0 20.0 136.65 MB

python Convention Gait Model

Home Page: http://cadop.github.io/pyCGM

License: MIT License

Python 100.00%

biomechanical-analysis biomechanics gait-analysis kinematics motion-capture python conventional-gait-model

pycgm's Introduction

pyCGM

Python Module for the Conventional Gait Model calculates Kinematics and Center of Mass

The goal of this project is to release an easy to understand conventional gait model that users can implement in their own projects via a single python file. While the project is multiple files, the kinematics code is contained in a single file pyCGM.py. With the update, another kinetics file is also used for the center of mass calculation.

Our aim is to provide researchers and students a tool that can aid in understanding and developing modifications to the conventional gait model through python without much more.

The kinematics are validated against Nexus 1.8, and file types are known from Nexus 1.8. Newer C3D files may not work (but files re-exported from Mokka usually work).

How to use?

For getting started, please check the WIKI on github

Requirements:

Python 2.7 or Python 3
Numpy
Scipy (only if using the pipeline operations)

Requirements for HPC:

Python 2.7
Numpy
MPI Preferably Linux (MPI is not as simple to setup on windows)

Uses a modified version of the c3d.py loader from github. https://pypi.python.org/pypi/c3d/0.2.1

Credits:

Originally developed in the Digital Human Research Center at the Advanced Institutes of Convergence Technology (AICT), Seoul National University http://aict.snu.ac.kr

Project Lead: Mathew Schwartz (umcadop at gmail) For issues, use github or email me directly

Contributors: Neil M. Thomas, Philippe C. Dixon, Seungeun Yeon (연승은),Filipe Alves Caixeta, Robert Van-Wesep

Reference

Read about this code and if you find it useful in your work please cite:

Schwartz, Mathew, and Philippe C. Dixon. "The effect of subject measurement error on joint kinematics in the conventional gait model: Insights from the open-source pyCGM tool using high performance computing methods." PloS one 13.1 (2018): e0189984. https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0189984

Bibtex:

@article{schwartz2018effect, title={The effect of subject measurement error on joint kinematics in the conventional gait model: Insights from the open-source pyCGM tool using high performance computing methods}, author={Schwartz, Mathew and Dixon, Philippe C}, journal={PloS one}, volume={13}, number={1}, pages={e0189984}, year={2018}, publisher={Public Library of Science} }

pycgm's People

Contributors

Stargazers

Watchers

pycgm's Issues

Python 2.7 build failing in all workflows

Currently, the Python 2.7 build (build2) in the GitHub workflow is failing automatically for all commits because of a handful on compatibility issues. The earliest of these examples is located in setup.py, where the builtin function open() has a keyword argument "encoding" that only exists for Python 3 and not Python 2. There are a few more issues in addition to this one, such as the python_requires setup option which is also in setup.py, and several doctests which have some sort of version difference between the Python builds.

Marker size is hardcoded

The marker size, while seemingly always 14mm, should be provided as a measurement with a default of 14 instead.

Existing Docstrings are not in proper Numpy format

The following functions from each of these files need to have their docstrings reformatted:

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/pyCGM.py

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/pycgm_embed.py

getfilenames
loadData
main

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/pycgmKinetics.py

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/pyCGM_Helpers.py

getfilenames

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/Pipelines.py

prep
clearMarker
filtering

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/pycgmStatic.py

normDiv in pyCGM_Single/pyCGM.py

Function normDiv in pyCGM_single/pyCGM.py should be calculation the normalization, vec, and then dividing every x, y, and z by vec. Instead it divides each x, y, z by vec twice. I believe just removing the following line,

pyCGM/pyCGM_Single/pyCGM.py

Line 2256 in 360fb86

def normDiv(v):

, should fix it:
v = [v[0]/vec,v[1]/vec,v[2]/vec]

The "Multiprocessing" folder is misspelled "Muliprocessing"

dataAsDict in pycgmIO.py has an improperly indented return statement.

def dataAsDict(data,npArray=False):
    dataDict = {}
    
    for frame in data:
        for key in frame:
            dataDict.setdefault(key,[])
            dataDict[key].append(frame[key])
    
    if npArray == True:
        for key in dataDict:
            dataDict[key] = np.array(dataDict[key])
        
        return dataDict

The return statement only executes when the boolean flag npArray is True because of its indentation. If npArray is False, the function returns nothing.

Relative imports fail when using doctest in docstrings

After having pip install the pyCGM module, a few of the relative imports do not work with doctests in docstrings.

For a example, a successful import in a doctest:

"""
>>>  from pycgmKinetics import unit
"""
if __name__ == "__main__":
    import doctest
    doctest.testmod()

Will yield the result:

Trying:
from pycgmKinetics import unit
Expecting nothing
Test passed.

While an doctest like this :

"""
>>> from pyCGM import cross
"""

if __name__ == "__main__":
    import doctest
    doctest.testmod()

Yields the error:

Failed example:
from pyCGM import cross
Exception raised:
Traceback (most recent call last):
File "C:\Users\GabrielVeras.windows-build-tools\python27\lib\doctest.py", line 1315, in __run
compileflags, 1) in test.globs
File "<doctest test file[0]>", line 1, in
from pyCGM import cross
File "pyCGM.py", line 30, in
from .pycgmIO import *
ValueError: Attempted relative import in non-package
Test Failed 1 failures.

In general, attempting to import from pyCGM or importing any method from the pyCGM file results in the same error.

Support for old c3d formats (ex. DEC integer format)

Maybe it would be great if the old c3d loader can be replaced by ezc3d (https://github.com/pyomeca/ezc3d), as it has support for a wider range of formats.

pyCGM_Single/pycgmClusters.py imports and uses a non-existent function

The file pycgmClusters.py contains one function, calcFramesClusters which imports a function from clusterCalc.py named groupInClustDict that does not exist.

On line 15, the function is imported:
from .clusterCalc import targetName, getMarkerLocation, targetDict, groupInClustDict ,getStaticTransform

On line 18, the function is used:
clusters = groupInClustDict()

Results are saved as "pycgm_results.csv.csv"

This is caused by the fact that pyCGM_Helpers.py, which supplies file names to whatever calls it, already has extensions within its file names. The function writeResult from pycgmIO.py appends an additional ".csv" to the filename that it accepted as a parameter.

Pep8 Name Formatting

At least try to do some pep8 formatting on functions.

Marker Names are Hardcoded

Currently marker names are hardcoded everywhere. This makes it hard to load a file that has, for example, a string appended subject:LFHD. It would be good to have a function that allows people to overwrite marker names easily. Perhaps this is done while loading the file as a mapping, or is done as a mapping after.

A pycgmStatic.py doctest results in a import math error

When running the command:
pytest --doctest-modules --continue-on-collection-errors --ignore=HPC/ --ignore=Muliprocessing/
in the main directory, I get the error:
UNEXPECTED EXCEPTION: NameError("name 'math' is not defined")
R_alpha = np.around(math.degrees(static_info[0][0]*-1),decimals=5)
NameError: name 'math' is not defined
specifically for a doctest that I created for the function footJointCenter in pycgmStatic.py. In my doctest I used both import math and from math import * and neither resolved this issue, even though import math should have.

Incorrect Syntax in pyCGM_Single/Pipelines.py

Minor incorrect syntax for the function prep in Pipelines.py.

When running the function prep with correct parameters, the following for loop generates an error:

 for key, val in trajs.iteritems():
                temp.update({key:val[i,:]})

The specific error raised:

    temp.update({key:val[i,:]})
TypeError: list indices must be integers or slices, not tuple

It is due to the slice being an incorrect format, instead i think it should just be temp.update({key:val[i:]}).
It is also worth noting that it is for both for loops in the function.

Setup Workflows

Setup the github workflows to test pull requests (although kind of needs doc and unit tests first).

Create a setup file

Make a setup file for pip install. Will also help with doctests dealing with relative imports.

Many functions are missing Docstrings

The following functions from each of these files need to have docstrings written for them in Numpy format:

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/clusterCalc.py

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/Pipelines.py

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/pycgmCalc.py

calcKinetics
calcAngles
Calc
calcFrames

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/pycgmClusters.py

calcFramesClusters

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/pycgmKinetics.py

getKinetics

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/pycgmIO.py

Some pyCGM.py unit tests still mock frame and vsk dictionaries

Many functions in pyCGM_Single/pyCGM.py used to take in:

frame: a dictionary of marker names to positions

 {
    "RTIB": np.array([433.97537231, 211.93408203, 273.3008728]),
    "LTIB": np.array([50.04016495,  235.90718079, 364.32226562]),
    "RANK": np.array([422.77005005, 217.74053955,  92.86152649]),
    "LANK": np.array([58.57380676,  208.54806519,  86.16953278]),
    ...
},

and vsk: a dictionary of subject measurement names to values

{
    "RightAnkleWidth": 70.0,
    "LeftAnkleWidth": 70.0,
    "RightTibialTorsion": 0.0,
    "LeftTibialTorsion": 0.0,
    ...
}

Some tests such as test_calc_axis_ankle still take in frame and vsk, but pull out the values:

pyCGM/pyCGM_Single/tests/test_pycgm_axis.py

Line 1880 in 78e10bb

rtib = frame["RTIB"] if "RTIB" in frame else None

before passing them directly to their respective functions:

pyCGM/pyCGM_Single/tests/test_pycgm_axis.py

Line 1896 in 78e10bb

    
           result = pyCGM.calc_axis_ankle(rtib, ltib, rank, lank, r_knee_jc, l_knee_jc, rank_width, lank_width, rtib_torsion, ltib_torsion)

While this is still a proper test, it would be more straightforward to properly parametrize the values instead of continuing to use the dictionaries.

new

Hi,
I come across your dev.
I dont know if your projet is still in progress.
If yes, i can say you i have a full-python code of the Vicon Plugin Gait Model.
Both kinematics and kinetics, were compared and verfified.

feel free to contact me if you were inerested in.

cheers

fabien

No Unit Tests have been implemented

The following functions from each of these files need to unit tests written for them:

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/clusterCalc.py

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/Pipelines.py

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/pyCGM.py

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/pycgmCalc.py

calcKinetics
calcAngles
Calc
calcFrames

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/pycgmClusters.py

calcFramesClusters

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/pycgmIO.py

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/pycgmKinetics.py

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/pycgmStatic.py

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/pycgm_embed.py

getfilenames
loadData
main

https://github.com/cadop/pyCGM/blob/master/pyCGM_Single/pyCGM_Helpers.py

getfilenames

Refactored pelvis axis unit test only accurate to 5 decimal places

The original test for pyCGM_Single/pyCGM.py/pelvisJointCenter expects the z component of the pelvis origin to equal 1032.89349365.

pyCGM/pyCGM_Single/tests/test_pycgm_axis.py

Line 966 in 1750a62

[np.array([251.60830688, 391.74131775, 1032.89349365]),

The new test for pelvisJointCenter (now pyCGM_Single/pyCGM.py/calc_axis_pelvis) expects the same value but in scientific notation: 1.03289349e+03

https://github.com/MattGonz/pyCGM/blob/15a9030e9c6e0835ab12f0e2bde0315488ce9b55/pyCGM_Single/tests/test_pycgm_axis.py#L973

It seems that when 1.03289349e+03 is evaluated and asserted almost equal to 1032.89349365 using np.testing.assert_almost_equal to six decimal places, the test fails.

Manual assertion: np.testing.assert_almost_equal(1032.89349365, 1032.89349, 6)

AssertionError:
Arrays are not almost equal to 6 decimals
 ACTUAL: 1032.89349365
 DESIRED: 1032.89349

loadC3D in pyCGM_Single/pyCGMIO.py does not reset the mydictunlabeled dictionary.

pyCGM/pyCGM_Single/pycgmIO.py

Line 133 in 360fb86

for frame_no, points, analog in reader.read_frames(True,True):

In the above loop, the dictionary mydict is set to an empty dictionary to prepare it for the next frame in the loop, but the dictionary mydictunlabeled is not. This causes unlabeled data to load incorrectly when using loadC3D.

cadop / pycgm Goto Github PK

pycgm's Introduction

pyCGM

How to use?

Requirements:

Requirements for HPC:

Credits:

Reference

Bibtex:

pycgm's People

Contributors

Stargazers

Watchers

Forkers

pycgm's Issues

Recommend Projects

Recommend Topics

Recommend Org