Skip to content

python_module

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

python_module

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
CMakeLists.txt
CMakeLists.txt
 
 
ImageStreamIOWrap.cpp
ImageStreamIOWrap.cpp
 
 
README.md
README.md
 
 

README.md

ImageStreamIOWrap

Installation

In the root directory ie. ImageStreamIO

pip install .

Note: You can install it in a custom path with the option -t $HOME/local/python

Note2: If you update it, use the option -U

Usage

import ImageStreamIOWrap as ISIO
img = ISIO.Image()

NOTE: img is empty at this time, you need to create or open a Image.

Creation of a new Image

In [2]: img.create?

Docstring:
create(*args, **kwargs)
Overloaded function.

1. create(self: ImageStreamIOWrap.Image, name: str, dims: numpy.ndarray[uint32], datatype: int=Type.FLOAT, shared: int=1, NBkw: int=1) -> int


            Create shared memory image stream
            Parameters:
                name     [in]:  the name of the shared memory file will be SHAREDMEMDIR/<name>_im.shm
                dims     [in]:  np.array of the image.
                datatype [in]:  data type code, pyImageStreamIO.Datatype
                shared   [in]:  if true then a shared memory buffer is allocated.  If false, only local storage is used.
                NBkw     [in]:  the number of keywords to allocate.
            Return:
                ret      [out]: error code


2. create(self: ImageStreamIOWrap.Image, name: str, dims: numpy.ndarray[uint32], datatype: int=Type.FLOAT, location: int=-1, shared: int=1, NBsem: int=10, NBkw: int=1, imagetype: int=2) -> int


            Create shared memory image stream
            Parameters:
                name      [in]:  the name of the shared memory file will be SHAREDMEMDIR/<name>_im.shm
                dims      [in]:  np.array of the image.
                datatype  [in]:  data type code, pyImageStreamIO.Datatype
                shared    [in]:  if true then a shared memory buffer is allocated.  If false, only local storage is used.
                NBsem     [in]:  the number of semaphores to allocate.
                NBkw      [in]:  the number of keywords to allocate.
                imagetype [in]:  type of the stream.
            Return:
                ret       [out]: error code

Type:      method

The first one is the legacy API where it uses CPU buffer.

The second one can create GPU IPC SHM (need CUDA).

In [3]: img.create("cacaoTest", [128,128], ISIO.ImageStreamIODataType.FLOAT, 1, 8)

Open of a existing Image

In [15]: img.open?

Docstring:
open(self: ImageStreamIOWrap.Image, name: str) -> int


Open / connect to existing shared memory image stream
Parameters:
    name   [in]:  the name of the shared memory file to connect
Return:
    ret    [out]: error code
Type:      method


In [16]: img.open("cacaoTest")

Use an Image

In [5]: img.md.size
Out[5]: [128, 128]

In [6]: img.semReadPID
Out[6]: [-1, -1, -1, -1, -1, -1, -1, -1, -1, -1]

In [7]: img.md.lastaccesstime
Out[7]: datetime.datetime(2019, 3, 10, 13, 16, 59, 516486)

In [8]: import numpy as np

In [9]: np.array(img)
Out[9]:
array([[0., 0., 0., ..., 0., 0., 0.],
       [0., 0., 0., ..., 0., 0., 0.],
       [0., 0., 0., ..., 0., 0., 0.],
       ...,
       [0., 0., 0., ..., 0., 0., 0.],
       [0., 0., 0., ..., 0., 0., 0.],
       [0., 0., 0., ..., 0., 0., 0.]], dtype=float32)

In [12]: img.write(np.ones((128,128), dtype=np.float32))

In [14]: np.array(img)
Out[14]:
array([[1., 1., 1., ..., 1., 1., 1.],
       [1., 1., 1., ..., 1., 1., 1.],
       [1., 1., 1., ..., 1., 1., 1.],
       ...,
       [1., 1., 1., ..., 1., 1., 1.],
       [1., 1., 1., ..., 1., 1., 1.],
       [1., 1., 1., ..., 1., 1., 1.]], dtype=float32)

In [15]: img.md.lastaccesstime
Out[15]: datetime.datetime(2019, 3, 10, 13, 18, 20, 132500)