Python for MATLAB Users

Promoting Open Source Computer Vision Research

Kitware, Inc. / Google Research

June 2012

Matt Leotta

Amitha Perera

Patrick Reynolds

Eran Swears

Yong Zhao

Varun Ganapathi

(Kitware / Google)

Python for MATLAB Users

June 2012

1 / 201

This presentation is copyrighted by Kitware, Inc. distributed under the Creative Commons by Attribution License 3.0 http://creativecommons.org/licenses/by/3.0 Source code used in this presentation and in accompanying examples and exercises is distributed under the Apache License 2.0 http://www.apache.org/licenses/LICENSE-2.0 Approved for Public Release, Distribution Unlimited

What this tutorial is about

Promote open source computer vision research.
Reduce the communitys dependence on MATLAB. Show MATLAB users how to migrate to Python. Present a Python IDE similar to that of MATLAB.

Introduce Python for scientic computing. Explain how to access functionality in Python from third party vision libraries (OpenCV, ITK). Cover some advanced computing topics using Python:
parallel processing with threads GPU processing with PyCUDA and PyOpenCL calling C/C++ code from Python calling legacy MATLAB code from Python

Provided examples and hands-on exercises for all of the above. Rae of an ASUS Transformer Pad and more.
(Kitware / Google) Python for MATLAB Users June 2012 3 / 201

Schedule (Approximate)
8:30 9:00 10:0010:30 10:30 11:15 12:301:30 1:30 2:15 3:003:30 3:30 4:15 4:45 5:00 Overview, Virtual Machine Setup Intro to Python Coee Break Python Scripting Scientic Python Numpy, Matplotlib, SciPy Lunch Using OpenCV Using Simple ITK Coee Break Parallel Processing / GPU Processing Interfacing Python with C/C++ Interfacing Python with MATLAB Wrap-up, Rae

(Kitware / Google)

Python for MATLAB Users

June 2012

4 / 201

Virtual Machine Distribution

Virtual Machine Distribution

Get a USB Memory Stick Install VirtualBox from it Import the virtual machine le Boot the virtual machine Get familiar with the environment and directories

(Kitware / Google)

Python for MATLAB Users

June 2012

6 / 201

Media Content
Directories and Files

VirtualBoxInstallers
VirtualBox-4.1.16-78094-Linux amd64.run (64-bit Linux) VirtualBox-4.1.16-78094-Linux x86.run (32-bit Linux) VirtualBox-4.1.16-78094-OSX.dmg (Mac) VirtualBox-4.1.16-78094-Win.exe (Windows) LinuxPackages
RPM (Fedora 1417, Mandriva 20102011, RHEL/CentOS 46, SUSE 1011, OpenSUSE 11.311.4) DEB (Debian 57, Ubuntu 8.04, 10.0412.04)

VirtualMachine
matlab-to-python.ova

https://www.virtualbox.org/wiki/Downloads (Kitware / Google) Python for MATLAB Users June 2012 7 / 201

Install VirtualBox

Select the installer for your platform Run it

(Kitware / Google)

Python for MATLAB Users

June 2012

8 / 201

Importing the Virtual Machine

Run VirtualBox In File Menu select Import Appliance Provide the lename in the USB stick VirtualMachine/matlab-to-python.ova You should see:

(Kitware / Google)

Python for MATLAB Users

June 2012

9 / 201

To Be Continued...

Why use Python?

Open Source Research

Sharing code and data is often overlooked but is as important as sharing algorithms and results in published scientic research. Shared code promotes reproducible research and scientic validation. Computer vision research code is commonly written in MATLAB. Sharing research code written in MATLAB is a good rst step, but . . . MATLAB code can only be executed by a privileged few who can aord MATLAB license fees.

(Kitware / Google)

Python for MATLAB Users

June 2012

12 / 201

Why not MATLAB?

Language design
MATLAB is eective for quick code and/or small code bases, but does not scale as projects grow Numerical computing is easy, but more general programing is not

Top-level functions and code organization

Function-name = le-name in MATLAB: discourages writing modular programs Hard to appropriately name functions and organize code without conicts

Toolboxes are purchased separately

Shared MATLAB code will only run if youve licensed the right toolboxes

Cost Very Expensive!

(Kitware / Google)

Python for MATLAB Users

June 2012

13 / 201

MATLAB Pricing
MATLAB Component Inidvidual Matlab License MATLAB Compiler Toolboxes Computer Vision System Image Processing Optimization Parallel Computing Statistics Neural Network Signal Processing Symbolic Math Total Pricing as of April 12th, 2012.
More Info: http://www.mathworks.com/store/productIndexLink.do (Kitware / Google) Python for MATLAB Users June 2012 14 / 201

Price $2,150 $5,000 $1,350 $1,000 $1,000 $1,000 $1,000 $1,000 $1,000 $1,000 $15,500 per user

Open Source Alternatives to MATLAB

Octave (GPL license)
A mostly MATLAB compatible language Command line only (no GUI IDE)

FreeMat (GPL license)

A somewhat MATLAB compatible language Very basic GUI

Scilab (CeCILL license)

Not compatible with MATLAB language, provides a converter Most full featured IDE

These options solve the cost and license restrictions, but have the same technical limitations MATLAB. None of these options are perfectly compatible with MATLAB code. Each has a relatively small user community.
(Kitware / Google) Python for MATLAB Users June 2012 15 / 201

Why Python?
Cost / License
Python and related packages are free (have no licensing fee) Python and related packages use permissive BSD-like licenses.

Language design
Python has well-designed language features supporting small and large code bases Python code is highly readable

Top-level functions and code organization

Python is modular and supports name spaces to avoid naming conicts.

Integration with other languages/tools An enormous user base.

Python is used for just about everything (web, database, networking, GUI, gaming, graphics, science, . . . )

Just search the web for testimonials

Interesting essay: http://www.stat.washington.edu/~hoytak/blog/whypython.html (Kitware / Google) Python for MATLAB Users June 2012 16 / 201

Ways to Use Python: Many IDE Choices

IDE Anjuta Eric Python IDE Geany IDLE Komodo IDE MonoDevelop NetBeans (7.0 and up will no longer support Python) Ninja_IDE PIDA PyCharm PyDev PyScripter Python Tools for Visual Studio Spyder Stani's Python Editor Wing wxGlade

Developer Naba Kumar Detlev Offenbach Team Guido van Rossum et al. ActiveState Novell and the Mono community Oracle Team Team JetBrains Aptana Microsoft Pierre Raybaut et al. Stani Wingware Alberto Griggio

Latest stable release version 3.0.3.0 4.4.18 and 5.1.6 0.20 3.2 6.0.0 2.4.2 6.9 2.0-Beta3 0.6.2 2.0.2 2.4.0 2.4.3 1.1 2.1.9 0.8.4h 4.1.3-1 0.6.3

Latest stable release date 2011-06-10 2011-10-02 2011-01-07 2011-05-13 2010-10-07 2011-01-19 2010-06-15 2010-08-04 2012-02-07 2012-02-01 2011-09-20 2012-02-10 2012-03-31 2008-02-14 2012-01-11 2008-02-02

Platform Unix-like Independent Independent Independent Cross-platform Cross-platform Cross-platform Independent Cross-platform Cross-platform Eclipse (Cross-platform) Windows Windows Independent Independent Linux/Windows/OS X Independent

Toolkit GTK+ Qt GTK2 Tkinter Mozilla platform Gtk# Swing Qt PyGTK Swing SWT

License GPL GPL GPL Open source Proprietary LGPL GPL GPL GPL Proprietary EPL MIT Licence Apache License 2.0

Qt wxPython GTK2 wxPython

MIT Licence GPL Proprietary MIT License

There are several choices for Python development environments. The above list is from Wikipedia (taken April 2012). Spyder is most like the MATLAB IDE.
http://en.wikipedia.org/wiki/List_of_integrated_devel...for_Python (Kitware / Google) Python for MATLAB Users June 2012 17 / 201

Ways to Use Python: IDLE Python IDE

The original IDE, bundled with most Python installations. IDLE is simple and not very intuitive to use.

More info: http://docs.python.org/library/idle.html (Kitware / Google) Python for MATLAB Users June 2012 18 / 201

Ways to Use Python: IPython Qt Console

IPython Qt Console allows for inline plotting. We will use IPython Qt Console to learn the basics of Python.
More info: http://ipython.org/ (Kitware / Google) Python for MATLAB Users June 2012 19 / 201

Ways to Use Python: SymPy with IPython Notebook

IPython Notebook lets you keep a math notebook in a web browser. SimPy adds symbolic math for a Maple or Mathematica alternative. Run ipython notebook --profile=sympy
More info: http://ipython.org/ and http://sympy.org/ (Kitware / Google) Python for MATLAB Users June 2012 20 / 201

Ways to Use Python: Spyder IDE

Spyder provides an IDE similar to MATLAB. We will use Spyder later in this tutorial.
More info: http://code.google.com/p/spyderlib/ (Kitware / Google) Python for MATLAB Users June 2012 21 / 201

Using the Virtual Machine

Starting the Ubuntu 12.04 VM

Select the matlab-to-python VM. Press the Start button. Wait for the OS to boot. You will be automatically logged in as username tutorial password tutorial

(Kitware / Google)

Python for MATLAB Users

June 2012

23 / 201

The Ubuntu 12.04 Desktop

The bar on the left is the App launcher. Some Apps have additional options when you right click. Drop down menus are on the top of the screen (as in MacOS X).

(Kitware / Google)

Python for MATLAB Users

June 2012

24 / 201

Key Applications

Key Applications
Firefox Spyder IPython Terminal A web browser. A MATLAB-like Python IDE. An enhanced interactive Python shell. The Gnome terminal emulator.

(Kitware / Google)

Python for MATLAB Users

June 2012

25 / 201

Introduction to Python

Playing with Python Interpreters

IPython is a Python interpreter with additional interactive capabilities and more MATLAB-like features, discussed later. We will start with IPython to learn the basics and then move to Spyder when working with Python script les. Note: both Python and IPython can be run in a terminal with no need for a GUI.
To test this, click in the launcher to open a terminal. Type python to start Python and ipython to start IPython. If your prompt is >>> you have a standard Python interpreter. If your prompt is In [1]: you have an IPython interpreter. Type quit() or Ctrl-D to exit, or close the terminal.

Python Interpreter
(Kitware / Google) Python for MATLAB Users

IPython Interpreter
June 2012 27 / 201

Launch the IPython Qt Console

Open the IPython Qt Console by clicking in the launcher.

(Kitware / Google)

Python for MATLAB Users

June 2012

28 / 201

Basic Math
Try typing commands directly into the IPython interpreter.
In [1]: 2 + 2 Out[1]: 4 In [2]: 2.0 + 2.0 Out[2]: 4.0 In [3]: 5.0 / 2 Out[3]: 2.5 In [4]: 5 / 2 Out[4]: 2 In [5]: 5.0 // 2.0 Out[5]: 2.0 In [6]: 5 % 2 Out[6]: 1 In [7]: 3 ** 2 Out[7]: 9 In [8]: _ + 2 Out[8]: 11 # This is a comment

# Floating point addition

# Floating point division

# Integer division ( in Python3 this returns 2.5)

# Double slash forces integer division

# 5 modulo 2 , in MATLAB this is mod (5 ,2)

# Exponentiation , in MATLAB this is 3^2

# Underscore is the last returned value , in MATLAB this is ans

More Examples: http://docs.python.org/tutorial/introduction.html (Kitware / Google) Python for MATLAB Users June 2012 29 / 201

Basic Built-in Types

Basic Numerical Types
int - Integer type with unlimited precision (in Python 3). float - Floating point type (usually double precision). complex - Complex number with real and imaginary oat components. Note: in Python 2 int has limited precision and separate type long has unlimited precision.

Special Types
bool - Boolean with values False or True. None - A null object, more about this later.

Container Types
str - A string of characters. tuple - A xed sequence of objects. list - A mutable sequence of objects. dict - A dictionary mapping one set of objects to another. Plus various lesser used containers.

Additional numerical types and containers are provided by NumPy.

More info: http://docs.python.org/library/stdtypes.html (Kitware / Google) Python for MATLAB Users June 2012 30 / 201

Type Conversions
Types are easily converted using the syntax type(object) Type conversion even works to and from strings Some type conversions (e.g. complex to oat) are not well dened and throw an exception.
In [9]: float (7) Out[9]: 7.0 In [10]: int (3.14159) Out[10]: 3 In [11]: str (28) Out[11]: 28 In [12]: float ( 5.5 ) Out[12]: 5.5 In [13]: complex (4.5) Out[13]: (4.5+0 j ) In [14]: bool (0) Out[14]: False In [15]: float ( True ) Out[15]: 1.0 More info: http://docs.python.org/library/stdtypes.html (Kitware / Google) Python for MATLAB Users June 2012 31 / 201

Assignment
Variables are created during assignment
In [16]: my_var -----------------------------------------------------------Traceback (most recent call last): File "<ipython console>", line 1, in <module> NameError: name my var is not defined In [17]: my_var = 10 In [18]: my_var Out[18]: 10

Other assignment examples

In [19]: x , y , z = 10 , 20 , 30 In [20]: ( y + z ) / x Out[20]: 5 In [21]: x += 5 In [22]: y *= 2 In [23]: x , y , z Out[23]: (15 , 40 , 30) More Examples: http://docs.python.org/tutorial/introduction.html (Kitware / Google) Python for MATLAB Users June 2012 32 / 201 # eqivalent to x = x + 5 # eqivalent to x = x * 2 # multiple assignment , uses tuples

Complex Numbers
Python natively supports complex numbers.
Imaginary numbers are written with the j or J sux.
In [24]: 1 j * 1 J Out[24]: ( -1+0 j ) In [25]: complex (3 , 2) Out[25]: (3+2 j ) In [26]: a = ( -1 + 3 j ) * 2 j In [27]: a Out[27]: ( -6 -2 j ) In [28]: a . real Out[28]: -6.0 In [29]: a . imag Out[29]: -2.0 In [30]: abs ( a ) Out[30]: 6 .3 2455 532 033 675 9 In [31]: a . conjugate () Out[31]: ( -6+2 j ) # access the real part # can also create with the complex function

# access the imaginary part

# abs ( a ) == sqrt ( a . real **2 + a . imag **2)

# the complex conjugate

More Examples: http://docs.python.org/tutorial/introduction.html (Kitware / Google) Python for MATLAB Users June 2012 33 / 201

Boolean Operators
Python x or y x and y not x MATLAB x || y x && y ~x Result if x is false, then y, else x (short-circuit) if x is false, then x, else y (short-circuit) if x is false, then True, else False Result if logical(x), then 1, else logical(y) (short-circuit) if logical(x), then logical(y), else 0 (short-circuit) if logical(x), then 0, else 1

Examples
MATLAB > 0.0 || hello 1 > 10 && (5 || 6) 1 > ~ 5 0 More info: http://docs.python.org/library/stdtypes.html (Kitware / Google) Python for MATLAB Users June 2012 34 / 201 Python > 0.0 or hello hello > 10 and (5 or 6) 5 > not 5 False

Comparisons
MATLAB < <= > >= == ~= N/A N/A Python < <= > >= == != is is not Meaning strictly less than less than or equal strictly greater than greater than or equal equal not equal object identity negated object identity

These comparisons always return a Boolean value. Comparison between dierent numeric types casts from int to float to complex as needed. For example 3.0 == 3 is True. Only not equal operators dier in syntax. The concept of object identity is not found in MATLAB.
More info: http://docs.python.org/library/stdtypes.html (Kitware / Google) Python for MATLAB Users June 2012 35 / 201

Object Identity
Every Python object has an identity. The function id(object) returns a unique identifying integer. The operation x is y is equivalent to id(x) == id(y). Assignment allows multiple variables to refer to the same object. For immutable types (e.g. numeric types, strings, tuples) creation of new objects may or may not refer to same-valued exiting object. For mutable type (e.g. list, dict) creation always creates a new object.
In [32]: x , y = 3 , 4.4 In [33]: x == y , x is y , id ( x ) , id ( y ) Out[33]: ( False , False , 30294792 , 32375168) In [34]: x = 2 + 2.4 In [35]: x == y , x is y , id ( x ) , id ( y ) # here x is y could be True or False Out[35]: ( True , False , 37311776 , 32375168) In [36]: x = y In [37]: x == y , x is y , id ( x ) , id ( y ) Out[37]: ( True , True , 32375168 , 32375168) # here x is y must be True

More info: http://docs.python.org/reference/datamodel.html (Kitware / Google) Python for MATLAB Users June 2012 36 / 201

Tuples
Tuples are immutable sequences of objects of any type. Tuple items are comma separated and often enclosed in parentheses. A trailing comma is required to make a tuple of length one.
In [38]: t = 1 , hello , (3.14 , None ) , In [39]: t [0] Out[39]: 1 2.5

# access item by index

In [40]: t [0] = 2 # item assignment not allowed ( immutable ) TypeError: tuple object does not support item assignment In [41]: t [1:3] # access a slice of the objects ( start : end ) Out[41]: ( hello , (3.14 , None )) In [42]: len ( t ) Out[42]: 4 In [43]: 2.5 in t Out[43]: True # the length of the tuple

# test for inclusion

In [44]: t + (5 ,) # concatenate tuples Out[44]: (1 , hello , (3.14 , None ) , 2.5 , 5) In [45]: (1 , 2) * 3 # concatenate n shallow copies Out[45]: (1 , 2 , 1 , 2 , 1 , 2) More info: http://docs.python.org/library/stdtypes.html (Kitware / Google) Python for MATLAB Users June 2012 37 / 201

Strings
Strings are immutable sequences of characters. Strings are enclosed in single or double quotes. The concatenation and indexing syntax is the same as Tuple. Strings have many built-in functions.
In [46]: s = " Hello World " In [47]: s . upper () Out[47]: HELLO WORLD In [48]: s [3:5] + - * 3 + s [:: -1] Out[48]: lo - - - dlroW olleH In [49]: s . split () Out[49]: [ Hello , World ] In [50]: - . join (( a , b , c )) Out[50]: a - b - c In [51]: s . replace ( " Hello " , " Goodbye " ) Out[51]: Goodbye World # to upper case

# concatenation and slicing

# string splitting

# join strings with a seperator

# substring replacement

In [52]: " item %05 d has value % g " % (23 , 3.14**2) Out[52]: item 00023 has value 9.8596 More info: http://docs.python.org/library/string.html (Kitware / Google)

# operator % works like sprintf

Python for MATLAB Users

June 2012

38 / 201

Lists
Lists are mutable sequences of objects of any type. List items are comma separated and enclosed in square brackets. The concatenation and indexing syntax is the same as Tuple.
In [53]: v = [1 , hello , (3.14 , None ) , In [54]: v [0] = 2 In [55]: v . append (0.5) In [56]: del v [2] 2.5]

# item assignment # append items to the end # delete an item

In [57]: v Out[57]: [2 , hello , 2.5 , 0.5] In [58]: v . sort () # sort items in place

In [59]: v Out[59]: [0.5 , 2 , 2.5 , hello ] In [60]: v = ([] ,)*3 In [61]: v [0]. append (1) In [62]: v Out[62]: ([1] , [1] , [1]) # a tuple with 3 shallow copies of the same empty list # append to the first list in the tuple # all lists update because they are shallow copies

More info: http://docs.python.org/library/stdtypes.html (Kitware / Google) Python for MATLAB Users June 2012 39 / 201

Dictionaries

Dictionaries are mutable mapping of one object to another. Key:value pairs are comma separated and enclosed in curly braces.
In [63]: d = { a : 1 , b : 2 , c : 3} In [64]: d [ a ] Out[64]: 1 In [65]: d [ d ] = 4 # access a value by key

# change a value or insert a new key : value

In [66]: d . keys () # a list of dictionary keys Out[66]: [ a , c , b , d ] In [67]: d . values () # a list of dictionary values Out[67]: [1 , 3 , 2 , 4] In [68]: d . items () # a list of ( key , value ) tuples Out[68]: [( a , 1) , ( c , 3) , ( b , 2) , ( d , 4)]

More info: http://docs.python.org/library/stdtypes.html (Kitware / Google) Python for MATLAB Users June 2012 40 / 201

IPython Commands
IPython adds several interactive capabilities to standard Python.
Tab Completion: start typing and hit the Tab key to get a list of possible completions. Works for variables, functions, paths, etc. History: access previous commands and their results
adds __ and ___ for second and third to last results. use the command history to list the history of commands. use In[n] to access the n-th command string. use Out[n] to access the n-th result.

Documentation: type objectName? to get info about objectName. Note, the standard Python help(objectName) also provides some documentation. Auto Call: lets you write functions as statements. For example, abs x calls abs(x). MATLAB also allows this. IPython provides many more features not covered here.

http://ipython.org/ipython-doc/rel-0.11/interactive/tutorial.html (Kitware / Google) Python for MATLAB Users June 2012 41 / 201

IPython Commands
Examples of using IPython history
In [1]: a = 1 In [2]: a += 1 In 1: 2: 3: [3]: history a = 1 a += 1 _ip . magic ( " history " )

In [4]: a Out[4]: 2 In [5]: In [2] Out[5]: u a += 1\ n In [6]: exec In [2] In [7]: a Out[7]: 3 In [8]: Out [4] + Out [7] Out[8]: 5 # like eval in MATLAB , Python also has eval for expressions

http://ipython.org/ipython-doc/rel-0.11/interactive/tutorial.html (Kitware / Google) Python for MATLAB Users June 2012 42 / 201

Python Scripting

Launch the Spyder IDE

So far weve been interacting with an interpreter. How do we write a script? Open the Spyder IDE by clicking in the launcher. Note: This version of Spyder (2.1.9) does not fully integrate with the installed version of IPython (0.12).
The standard Python shell in Spyder has been enhanced with some features of IPython, like tab completion. IPython will run within Spyder, but certain capabilities, like plotting, do not work. IPython integration is scheduled to be xed in Spyder 2.2, planned for October 2012. For now we will use the standard interpreter, the one with the >>>prompt.

(Kitware / Google)

Python for MATLAB Users

June 2012

44 / 201

Using the Spyder IDE

Spyder has an editor, interpreter, variable explorer, object inspector ...

(Kitware / Google)

Python for MATLAB Users

June 2012

45 / 201

Hello World
The simple way
print " Hello World "

Note: In Python 3 print is a function, not a statement.

print ( " Hello World " ) # Python 3 version

A more complete example Examples/HelloWorld.py

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 # !/ usr / bin / python # Copyright 2011 , Kitware , Inc .

""" Prints the string " Hello World " and exits . """ my_text = " Hello World " def main (): """ The main function . """ print my_text

if __name__ == " __main__ " : main ()

(Kitware / Google)

Python for MATLAB Users

June 2012

46 / 201

Hello World: line-by-line

Comments. The rst line is a shebang directive for Unix.
1 2 # !/ usr / bin / python # Copyright 2011 , Kitware , Inc .

Documentation string (docstring) for the script. Uses triple quotes.

5 """ Prints the string " Hello World " and exits . """

Assigns the Hello World string to a global variable.

7 my_text = " Hello World "

A function denition with docstring. It calls the print function.

9 10 11 def main (): """ The main function . """ print my_text

Call the main function only if this le is run as a script (not imported).
14 15 if __name__ == " __main__ " : main ()

(Kitware / Google)

Python for MATLAB Users

June 2012

47 / 201

Running a Script

In Spyder run the active script with Menu RunRun, or press F5 Adjust the conguration settings and press the Run button Future runs use the same conguration Change the run conguration with Menu RunCongure, or press F6 Alternatively, in a terminal window (from the home directory) run python Tutorial/Examples/HelloWorld.py

(Kitware / Google)

Python for MATLAB Users

June 2012

48 / 201

Conditionals
MATLAB if a > 10 a = 10; end Python if a > 10: a = 10

Syntax of basic conditionals is very similar. MATLAB uses keyword end to designate the end of a block. Python uses : to start a block and indentation to designate duration.
These rules apply generally to loops, functions, classes, etc. Indentation is important in Python! Choose any indentation, but consistent indentation denes a block.
MATLAB if a > 10 disp ( a greater than 10 ) elseif a < 5 disp ( a less than 5 ) else disp ( a is , a ) end Python if a > 10: print a greater than 10 elif a < 5: print a less than 5 else : print a is , a

Note, elseif in MATLAB becomes elif in Python.

(Kitware / Google) Python for MATLAB Users June 2012 49 / 201

Loops
MATLAB x = 1; while x < 10 x = x * 2; end Python x = 1 while x < 10: x = x * 2

MATLAB and Python while loops are essentially the same. break and continue statements are the same in both languages.
MATLAB X = [5 4 3 2 1]; for v = X disp ( v ) end Python X = [5 , 4 , 3 , 2 , 1] for v in X : print v

MATLAB and Python for loops each iterate over a container. MATLAB uses symbol = while Python uses keyword in.

(Kitware / Google)

Python for MATLAB Users

June 2012

50 / 201

Generating Ranges
MATLAB has special syntax to generate a range of numbers
first:last first:step:last

Python use the range function to achieve similar behavior

range(end) range(first, end) range(first, end, step)
MATLAB > 0:4 0 1 2 > 2:5 2 3 4 > 6: -2:2 6 4 2 Python > range (5) [0 , 1 , 2 , 3 , 4] > range (2 , 6) [2 , 3 , 4 , 5] > range (6 , 1 , -2) [6 , 4 , 2]

3 5

Note that last is inclusive while end is not. In MATLAB first i last, but in Python first i <end. Pythons range function is limited to integer types. NumPy adds the arange function which generalizes to oats.
(Kitware / Google) Python for MATLAB Users June 2012 51 / 201

Dening Functions
MATLAB function [s , d ] = sum_diff (x , y ) % SUM_DIFF Sum and diff of two values . % x is the first value % y is the second value s = x + y; d = x - y; Python def sum_diff (x , y ): """ Sum and diff of two values . x is the first value y is the second value """ return x + y , x - y

MATLAB handles return values by explicitly naming them.

returns zero, one, or multiple named variables.

Python always returns exactly one object with the return statement.
a standalone or missing return statement returns None. returning multiple comma-separated objects (as above) returns a tuple.
MATLAB > [ v1 , v2 ] = sum_diff (7 , 4) v1 = 11 v2 = 3 > v3 = sum_diff (7 , 4) v3 = 11 Python > v1 , v2 = sum_diff (7 , 4) > v3 = sum_diff (7 , 4) > print " v1 = " , v1 , " , v2 = " , v2 , \ " , v3 = " , v3 v1 = 11 , v2 = 3 , v3 = (11 , 3)

(Kitware / Google)

Python for MATLAB Users

June 2012

52 / 201

Dening Functions Optional Arguments

MATLAB function arguments are all optional, but you can enforce the number of arguments and set default values by checking nargin. Python function arguments are required unless a default value is specied. All optional arguments must come after all required arguments.
MATLAB function val = my_func ( req , opt1 , opt2 ) if nargin < 3 opt2 = default string ; elseif nargin < 2 opt1 = 3.14; end disp ( opt2 ) val = req * opt1 Python def my_func ( req , opt1 =3.14 , opt2 = default string ): print opt2 return req * opt1

Python optional arguments can be omitted and/or specied as key=value pairs in any order, e.g in the above example my_func(1, 3.14, hi) is the same as my_func(1, opt2=hi)
http://docs.python.org/tutorial/controlflow.html#keyword-arguments (Kitware / Google) Python for MATLAB Users June 2012 53 / 201

Managing Libraries of Functions

MATLAB and Python have dierent strategies for organizing and calling libraries of functions. In MATLAB, importing functions is implicit.
Each function must reside in a le of the same name. When a function is called, the MATLAB path is searched to nd the rst matching function. Dealing with functions in dierent packages with the same name is problematic.

In Python, importing functions is explicit.

Any number of functions can reside in a le. Modules, packages, and function must be explicitly imported when needed. Name spaces and renaming allows for easy resolution of name conicts.

(Kitware / Google)

Python for MATLAB Users

June 2012

54 / 201

Importing Packages, Modules, Functions, etc.

The Python import statement is used to explicitly import packages, modules, classes, functions, and variables. module is a python le (.py) that may contain multiple functions, variables, and classes. package is a collection of modules in a directory. The directory must contain a le named init .py. Packages can be nested. The import statement initializes a module or package by executing the code in the module or init .py. Python searches current directory and Python path to nd matching modules and packages. Each module or package is imported only once in a Python session, call reload() to force reloading.
http://docs.python.org/tutorial/modules.html (Kitware / Google) Python for MATLAB Users June 2012 55 / 201

An Example Python Package

Directory structure of a Python package
vision/ __init__.py learning/ __init__.py adaboost.py svm.py tracking/ __init__.py kalman.py features/ __init__.py sift.py harris.py canny.py
http://docs.python.org/tutorial/modules.html (Kitware / Google) Python for MATLAB Users June 2012 56 / 201

Top-level package Initialize vision package Subpackage for learning

Subpackage for tracking

Subpackage for features

Importing Modules and Packages

If the vision directory is in the Python path then the following works.
import vision . tracking . kalman vision . tracking . kalman . predict () # executes vision / tracking / kalman . py # call a function in kalman . py

Generally init .py includes statements to import all the sub-packages or modules recursively. So the following usually works, but it imports the whole vision package.
import vision vision . features . harris . detect () vision . tracking . kalman . predict () # executes vision / __init__ . py # call a function in harris . py # call a function in kalman . py

The as keyword allows renaming during import.

import vision . tracking . kalman as kf kf . predict ()

http://docs.python.org/tutorial/modules.html (Kitware / Google) Python for MATLAB Users June 2012 57 / 201

Importing Specic Functions or Variables

The from keyword allows importing into the current name space.
from vision . tracking . kalman import predict , correct predict () correct ()

Rename specic functions with as.

from vision . tracking . kalman import predict as my_pred , correct as my_corr my_pred () my_corr ()

Using * imports of everything into the current name space.

from vision . tracking . kalman import * predict () correct ()

http://docs.python.org/tutorial/modules.html (Kitware / Google) Python for MATLAB Users June 2012 58 / 201

The Python Path

The Python path is set with environment variable PYTHONPATH. Modify sys.path to make temporary changes during execution.
In [9]: import sys In [10]: sys . path . append ( " / home / tutorial " ) In [11]: sys . path Out[11]: [, / usr / bin , / usr / lib / python2 .7 , ... / usr / lib / python2 .7/ dist - packages / IPython / extensions , / home / tutorial ]

http://docs.python.org/tutorial/modules.html (Kitware / Google) Python for MATLAB Users June 2012 59 / 201

Exercise 1
Basic Python

Open the le Exercises/ex1.py Write the missing function sqrt_dict to compute a dictionary mapping the rst n integers to the square roots of their values. Hint: use the sqrt function from the math module. This example uses sys.argv to access a list of command line arguments. Exercises/ex1.py
19 20 21 22 23 24 25 26 27 28 # parse the command line argument # require exactly 1 argument ( argv [0] is the script name ) if len ( sys . argv ) != 2: print " usage : % s number " % sys . argv [0] return # convert the command line argument to an interger num = int ( sys . argv [1]) # call the function values = sqrt_dict ( num )

(Kitware / Google)

Python for MATLAB Users

June 2012

60 / 201

Exercise 1: Answer
Basic Python

Here is one possible answer Exercises/Answers/ex1-answer.py

6 7 8 9 10 11 12 13 14 15 import sys import math

def sqrt_dict ( n ): """ Compute a dictionary mapping the first n integers to square roots . """ d = dict () for i in range ( n ): d [ i ] = math . sqrt ( i ) return d

An alternate one-line solution uses list comprehensions Exercises/Answers/ex1-answer.py

22 return dict ([( i , math . sqrt ( i )) for i in range ( n )])

(Kitware / Google)

Python for MATLAB Users

June 2012

61 / 201

Scientic Computing in Python

NumPy, SciPy, and Matplotlib

NumPy adds basic MATLAB-like capability to Python:
multidimensional arrays with homogeneous data types specic numeric data types (e.g. int8, uint32, float64) array manipulation functions (e.g. reshape, transpose, concatenate) array generation (e.g. ones, zeros, eye, random) element-wise math operations (e.g. add, multiply, max, sin) matrix math operations (e.g. inner/outer product, rank, trace) linear algebra (e.g. inv, pinv, svd, eig, det, qr)

SciPy builds on NumPy (much like MATLAB toolboxes) adding:

multidimensional image processing non-linear solvers, optimization, root nding signal processing, fast Fourier transforms numerical integration, interpolation, statistical functions sparse matrices, sparse solvers clustering algorithms, distance metrics, spatial data structures le IO (including to MATLAB .mat les)

Matplotlib adds MATLAB-like plotting capability on top of NumPy.

(Kitware / Google) Python for MATLAB Users June 2012 63 / 201

Interactive Scientic Python (aka PyLab)

PyLab is a meta-package that includes NumPy, SciPy, and Matplotlib. The easiest (and most MATLAB-like) way to work with scientic Python is to import everything into the global name space.
from pylab import *

IPython can be started with the --pylab command line option. Spyder starts its standard Python interpreter with these scientic libraries imported, much like PyLab, but not PyLab per se.

More info: http://www.scipy.org/PyLab (Kitware / Google) Python for MATLAB Users June 2012 64 / 201

Scripting with Scientic Python

When writing scripts it is recommended that you:
only import what you need, for eciency import packages into namespaces, to avoid name clashes

The community has adopted abbreviated naming conventions:

import import import import numpy as np scipy as sp matplotlib as mpl matplotlib . pyplot as plt

Some dierent ways of working with NumPy are

from pylab import * svd ( eye (3)) from numpy import eye , array from numpy . linalg import svd svd ( eye (3)) import numpy as np np . linalg . svd ( np . eye (3)) # Import everything . Good for interactive work , # but not recommended for writing scripts . # Import only what you need into the global namespace . # Verbose syntax for import statements , # but simple syntax for calling functions . # Import a superset of what you need into a namespace . # More verbose syntax for calling functions .

http://docs.scipy.org/doc/scipy/reference/tutorial/general.html (Kitware / Google) Python for MATLAB Users June 2012 65 / 201

NumPy

Numpy provides a good website for mapping MATLAB syntax into equivalent NumPy syntax: http://www.scipy.org/NumPy_for_Matlab_Users This tutorial will only cover the most common cases.

More info: http://www.scipy.org/NumPy_for_Matlab_Users (Kitware / Google) Python for MATLAB Users June 2012 66 / 201

NumPy Array
Use numpy.array to represent multidimensional arrays.
>>> A = array ([1 , 2 , 3]) >>> B = array ([[1 , 2 , 3] , [4 , 5 , 6]]) # create a one dimensional array from a list

# create a two dimensional array from a list of lists

>>> C = array ([ B , B , B , B ]) # create a 3 D array from a list of 2 D arrays >>> A . shape (3 ,) >>> B . shape (2 ,3) >>> C . shape (4 ,2 ,3) >>> A . ndim 1 >>> A . dtype dtype ( int64 ) # the number of dimensions # the size of each dimension

# the data type of each element

(Kitware / Google)

Python for MATLAB Users

June 2012

67 / 201

NumPy Array

Unlike MATLAB, operations on numpy.arrays are element-wise.

MATLAB > M = [1 2; 2 1]; > disp ( M * M ) 5 4 4 5 > disp ( M .* M ) 1 4 4 1 Python > M = array ([[1 , 2] ,[2 , 1]]) > print M * M [[1 4] [4 1]] > print dot (M , M ) [[5 4] [4 5]]

The function numpy.dot is required for matrix (or tensor) multiplication. This syntax may seem ugly, but makes operations more consistent over arrays with dimensions greater than 2.

(Kitware / Google)

Python for MATLAB Users

June 2012

68 / 201

NumPy Matrix

Numpy provides a special numpy.matrix subclass of numpy.array. numpy.matrix acts more like a MATLAB matrix.
MATLAB > M = [1 2; 2 1]; > disp ( M * M ) 5 4 4 5 Python > M = matrix ( 1 2; 2 1 ) > print M * M [[5 4] [4 5]]

numpy.matrix can use a MATLAB-style string initializer. Use from numpy.matlib import * to override common array constructors (zeros(), eye(), etc.) with matrix versions. numpy.matrix only handles two dimensional arrays. numpy.array is more consistent and still preferred in most cases.

More info: http://www.scipy.org/NumPy_for_Matlab_Users (Kitware / Google) Python for MATLAB Users June 2012 69 / 201

NumPy Numerical Data Types

Numpy adds several numerical data types to Python
Names bool int int8, int16, int32, int64 uint8, uint16, uint32, uint64 float float16, float32, float64 complex complex64, complex128 Descriptions Boolean (True or False) stored in a byte Platform integer (usually int32 or int64) 8, 16, 32 and 64 bit signed integers 8, 16, 32 and 64 bit unsigned integers Shorthand for float64 half, single, and double precision oats Shorthand for complex128 complex numbers with single and double precision oats for real and imaginary parts

Type can be specied during array creation and converted later

>>> A = np . array ([1 , 2 , 3] , dtype = np . float32 ) >>> A array ([ 1. , # construct as type float32

2. ,

3.] , dtype = float32 ) # convert the type of an existing array 3.+0. j ])

>>> A . astype ( complex ) array ([ 1.+0. j , 2.+0. j ,

More info: http://docs.scipy.org/doc/numpy/user/basics.types.html (Kitware / Google) Python for MATLAB Users June 2012 70 / 201

NumPy Array Construction

Common helper functions for constructing arrays are similar to MATLAB.

MATLAB > > > > > > > M M M M M M M = = = = = = = zeros (3 ,4); ones (3 ,4); eye (3); diag ([1 2 3]); diag ( A ); rand (3 ,4); repmat (A , m , n ); Python > > > > > > > M M M M M M M = = = = = = = zeros ((3 ,4)) ones ((3 ,4)) eye (3) diag ([1 , 2 , 3]) diag ( A ) random . rand (3 ,4) tile (A , (m , n ))

More info: http://www.scipy.org/NumPy_for_Matlab_Users (Kitware / Google) Python for MATLAB Users June 2012 71 / 201

NumPy Array Slicing

Slicing allow access to subset of arrays. Python slicing syntax is first:end:step.

MATLAB a(end) a(2,5) a(2,:) a(1:5,:) a(3:2:21,:) a(1:2:end,:) a(end:-1:1,:) NumPy a[-1] a[1,4] a[1] or a[1,:] a[0:5] or a[:5] or a[0:5,:] a[ 2:21:2,:] a[ ::2,:] a[ ::-1,:] Description access last element in the 1xn matrix a. access element in second row, fth column. access entire second row of a. access the rst ve rows of a. every other row of a, from the third to the twenty-rst. every other row of a, starting with the rst. a with rows in reverse order.

More info: http://www.scipy.org/NumPy_for_Matlab_Users (Kitware / Google) Python for MATLAB Users June 2012 72 / 201

NumPy Indexing Tricks Stacking

Numpy provides a special class to enable several indexing tricks. The r_ class provides row stacking of arrays r_ is used with brackets as if it is being indexed.
>>> r_ [[1 , 2 , 3] , [4 , 5 , 6] , [7 , 8]] array ([1 , 2 , 3 , 4 , 5 , 6 , 7]) >>> r_ [ eye (2) , 2* ones ((2 ,2)) ] array ([[ [ [ [ 1. , 0. , 2. , 2. , 0.] , 1.] , 2.] , 2.]]) # stack 1 D arrays along the first axis

# stack 2 D arrays along the first axis

A variant named c_ is same, but for column stacking.

>>> c_ [ eye (2) , 2* ones ((2 ,2)) ] array ([[ 1. , [ 0. , 0. , 1. , 2. , 2. , 2.] , 2.]]) # stack 2 D arrays along the second axis

More info: http://www.scipy.org/NumPy_for_Matlab_Users (Kitware / Google) Python for MATLAB Users June 2012 73 / 201

NumPy Indexing Tricks Range Generation

The r_ class also provides range generation. If an argument uses slice notation it is expanded to a range r_[first:end:step] is the same as arange(first,end,step)
>>> r_ [5:10:0.5] array ([ 5. , 5.5 , # generates a range from 5 to 10 with step 0.5 6. , 6.5 , 7. , 7.5 , 8. , 8.5 , 9. , 9.5])

When the step is replaced with an imaginary number the slice has a dierent interpretation, first:last:num
>>> r_ [5:10:5 j ] array ([ 5. , # generates 5 samples between 5 and 10 , inclusive 6.25 , 7.5 , 8.75 , 10. ])

More info: http://www.scipy.org/NumPy_for_Matlab_Users (Kitware / Google) Python for MATLAB Users June 2012 74 / 201

NumPy Indexing Tricks

The ix_ function generates indexes for irregular slices

MATLAB a([2,4,5],[1,3]) NumPy a[ix_([1,3,4],[0,2])] Description rows 2,4 and 5 and columns 1 and 3.

More info: http://www.scipy.org/NumPy_for_Matlab_Users (Kitware / Google) Python for MATLAB Users June 2012 75 / 201

Matplotlib
d + v v = p + 2 v + g dt x2 W311 2 = U311 + 81 2

dx =

matplotlib
2

m1 m2 FG = G r2
2

d2

U211 2 U112 U012

Matplotlib provides plotting capabilities that are similar to MATLAB. Syntax is the same or similar to MATLAB in most cases. Matplotlib has some advantages over MATLAB plots
Handles TeX equation typesetting within plots. Easy export to PDF using vector graphics. Plots often look nicer than MATLAB plots.

The above logo is actually a matplotlib plot.

More info: http://matplotlib.sourceforge.net/ (Kitware / Google) Python for MATLAB Users June 2012 76 / 201

Matplotlib Example Plots

1.0 0.5 voltage (mV) 0.0 0.5 1.0 0.0 40 35 30 25 Scores 20 15 10 5 0 G1 G2 G3 G4 G5 4 2 0 2 4 4 0 2 2 20 25 20 Probability

About as simple as it gets, folks

0.030 0.025 0.020 0.015 0.010 0.005

Histogram of IQ : =100, =15

0.5

1.0 time (s)

1.5

2.0

0.000 40

60

80

100 120 Smarts

140

160

Scores by group and gender 35 Men 34 35 32 Women 30

27 25

1.01 0.79 0.56 0.34 0.11 -0.11 -0.34 -0.56 -0.79 -1.01 4

0.8 0.6 0.4 0.2 0.0 0.2 0.4 0.6 0.8

More info: http://matplotlib.sourceforge.net/users/screenshots.html (Kitware / Google) Python for MATLAB Users June 2012 77 / 201

Matplotlib Example Plots

Examples/simple plot.py
1 2 3 4 5 6 7 8 9 10 11 from pylab import * t = arange (0.0 , 2.0 , 0.01) s = sin (2 * pi * t ) plot (t , s , linewidth =1.0) xlabel ( time ( s ) ) ylabel ( voltage ( mV ) ) title ( About as simple as it gets , folks ) grid ( True ) show ()

When not using pylab, these functions are found in matplotlib.pyplot. The gure is not displayed until show() is called.

More info: http://matplotlib.sourceforge.net/users/screenshots.html (Kitware / Google) Python for MATLAB Users June 2012 78 / 201

Matplotlib Interactive Plotting

Matplotlib has an interactive plotting mode that updates plots after each command, and show() is not needed. Interactive plotting is enabled with ion() and disabled with ioff(). Interactive mode is enabled automatically in Spyder and ipython --pylab

More info: http://matplotlib.sourceforge.net/users/shell.html (Kitware / Google) Python for MATLAB Users June 2012 79 / 201

Exercise 2
Plotting Lena with NumPy, SciPy, and Matplotlib
1 2 3

Plot the Lena image.

>>> from scipy . misc import lena >>> imshow ( lena ()) >>> gray ()

Plot the transpose of Lena using lena().T. Plot 100 random points on top of Lena with x and y between 0 and 512 using rand and scatter. Plot a histogram of Lena with 100 bins using hist. Hint: Flatten the image to 1-D using lena().flatten() Plot Lena with a sub-matrix from (200,200) to (400,400) ipped horizontally. Hint: Create a modied image array using slice notation.

Part 1
(Kitware / Google)

Part 2

Part 3
Python for MATLAB Users

Part 4

Part 5
June 2012 80 / 201

Exercise 2: Answers
Plotting Lena with NumPy, SciPy, and Matplotlib

Exercises/Answers/ex2-answer.py
7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 from scipy . misc import lena figure ( " Part 1 " ) imshow ( lena ()) gray () figure ( " Part 2 " ) imshow ( lena (). T ) gray () figure ( " Part 3 " ) imshow ( lena ()) points = rand (2 , 100) * 512 scatter ( points [0] , points [1]) figure ( " Part 4 " ) hist ( lena (). flatten () , 100) figure ( " Part 5 " ) image = lena () image [200:400 , 200:400] = image [200:400 , 400:200: -1] imshow ( image )

(Kitware / Google)

Python for MATLAB Users

June 2012

81 / 201

SciPy

Contains many scientic Python modules built on top of NumPy. Most interesting for computer vision work are ndimage Multi-dimensional image processing linalg Linear Algebra (extends numpy.linalg) tpack Fourier Transforms optimize Optimization io File IO (read/write MATLAB les) In this tutorial, we are only going to scratch the surface of what is contained in these libraries with a few examples.

More info: http://www.scipy.org/ (Kitware / Google) Python for MATLAB Users June 2012 82 / 201

SciKits

There are also many add-on toolkits for SciPy, collectively called SciKits Most interesting for computer vision work are scikits-image Image Processing scikits-learn Machine Learning and Data Mining ann Approximate Nearest Neighbor library wrapper mlabwrap Call MATLAB from Python

More info: http://scikits.appspot.com/ (Kitware / Google) Python for MATLAB Users June 2012 83 / 201

SciPy Central

Looking for an equivalent to MATLAB Central for sharing snippets of scientic Python code? Try SciPy Central (scipy-central.org) Its not as extensive as MATLAB Central, but its fairly new.

More info: http://scipy-central.org/ (Kitware / Google) Python for MATLAB Users June 2012 84 / 201

SciPy By Example

Lets work through an example of writting a Canny edge detector. This is not a very ecient implementation of Canny edge detection. The goal is to illustrates some SciPy and Numpy functionality. Open the le Examples/scipy canny.py

(Kitware / Google)

Python for MATLAB Users

June 2012

85 / 201

SciPy By Example Importing

First import what we need: NumPy and and NDImage from SciPy. These modules are imported to abreviated names np and ndi. Note that the constant pi (i.e. ) is imported into the global namespace to make math expressions easier to follow. Examples/scipy canny.py
4 5 6 import numpy as np from numpy import pi import scipy . ndimage as ndi

(Kitware / Google)

Python for MATLAB Users

June 2012

86 / 201

SciPy By Example Main

Now look at the end of the le. Examples/scipy canny.py
46 47 48 49 50 51 if __name__ == " __main__ " : from scipy . misc import lena from matplotlib . pyplot import imshow , gray , show imshow ( canny_edges ( lena ())) gray () show ()

Usually all imports happen at the top of a le, but this is not required. In this example some more things are imported only when this le is run as a script. When run as a script, this code calls canny_edges on the Lena image and displays the result in a matplotlib GUI. When imported into another module, there is no dependency on matplotlib or Lena.
(Kitware / Google) Python for MATLAB Users June 2012 87 / 201

SciPy By Example Result

Calling imshow(lena()) would show the input image on the left. Running this script displays the image on the right.

Input Image

Canny Edges

(Kitware / Google)

Python for MATLAB Users

June 2012

88 / 201

SciPy By Example Canny

Here is the Canny edge detection code. Examples/scipy canny.py
27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 def canny_edges ( image , sigma =1.0 , low_thresh =50 , high_thresh =100): """ Compute Canny edge detection on an image . """ image = ndi . filters . gaussian_filter ( image , sigma ) dx = ndi . filters . sobel ( image , 0) dy = ndi . filters . sobel ( image , 1) mag = np . sqrt ( dx ** 2 + dy ** 2) ort = np . arctan2 ( dy , dx ) edge_map = n o n _ m a x i m a l _ e d g e _ s u p p r e s i o n ( mag , ort ) edge_map = np . logical_and ( edge_map , mag > low_thresh ) labels , num_labels = ndi . measurements . label ( edge_map , np . ones ((3 , 3))) for i in range ( num_labels ): if max ( mag [ labels == i ]) < high_thresh : edge_map [ labels == i ] = False return edge_map

Only non_maximal_edge_suppression is not provided by np or ndi.

(Kitware / Google) Python for MATLAB Users June 2012 89 / 201

SciPy By Example Gaussian Filter

1

Smooth the image with a Gaussian Filter. Examples/scipy canny.py

29

image = ndi . filters . gaussian_filter ( image , sigma )

Input Image
(Kitware / Google)

Gaussian Filter with = 5

Python for MATLAB Users June 2012 90 / 201

SciPy By Example Sobel

2

Sobel derivatives on smoothed image. Examples/scipy canny.py

30 31

dx = ndi . filters . sobel ( image , 0) dy = ndi . filters . sobel ( image , 1)

dx Sobel horizontal
(Kitware / Google) Python for MATLAB Users

dy Sobel vertical
June 2012 91 / 201

SciPy By Example Gradient Mag/Orient

3

Sobel derivatives on smoothed image. Examples/scipy canny.py

33 34

mag = np . sqrt ( dx ** 2 + dy ** 2) ort = np . arctan2 ( dy , dx )

Gradient Magnitude
(Kitware / Google) Python for MATLAB Users

Gradient Orientation
June 2012 92 / 201

SciPy By Example Non-Max Suppression

4

Non-maximal suppression (more on this later). Examples/scipy canny.py

36

edge_map = n o n _ m a x i m a l _ e d g e _ s u p p r e s i o n ( mag , ort )

All maximimal edges (binary image)

(Kitware / Google) Python for MATLAB Users June 2012 93 / 201

SciPy By Example Threshold

5

Apply the lower threshold Examples/scipy canny.py

37

edge_map = np . logical_and ( edge_map , mag > low_thresh )

Low Threshold (50)

(Kitware / Google) Python for MATLAB Users

High Threshold (100)

June 2012 94 / 201

SciPy By Example Label Connected Components

6

Label connected components using 8-neighborhood (3x3 mask of ones). Examples/scipy canny.py
labels , num_labels = ndi . measurements . label ( edge_map , np . ones ((3 , 3)))

39

Connected edges with random colors

(Kitware / Google) Python for MATLAB Users June 2012 95 / 201

SciPy By Example Hysteresis

7

Remove connected edges not containing a value above the high threshold. Examples/scipy canny.py

40 41 42

for i in range ( num_labels ): if max ( mag [ labels == i ]) < high_thresh : edge_map [ labels == i ] = False

Low Threshold (50)

(Kitware / Google)

Hysteresis
Python for MATLAB Users

High Threshold (100)

June 2012 96 / 201

SciPy By Example Non-Max Suppression

Non-Maximal Suppression of edges produces a binary edge map. Gradient magnitude is a maximum in the direction normal to an edge. Examples/scipy canny.py
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 def n o n _ m a x i m a l _ e d g e _ s u p p r e s i o n ( mag , orient ): """ Non Maximal suppression of gradient magnitude and orientation . """ # bin orientations into 4 discrete directions abin = (( orient + pi ) * 4 / pi + 0.5). astype ( int ) % 4 mask = np . zeros ( mag . shape , dtype = bool ) mask [1: -1 , 1: -1] = True edge_map = np . zeros ( mag . shape , dtype = bool ) offsets = ((1 , 0) , (1 , 1) , (0 , 1) , ( -1 , 1)) for a , ( di , dj ) in zip ( range (4) , offsets ): cand_idx = np . nonzero ( np . logical_and ( abin == a , mask )) for i , j in zip (* cand_idx ): if mag [i , j ] > mag [ i + di , j + dj ] and \ mag [i , j ] > mag [ i - di , j - dj ]: edge_map [i , j ] = True return edge_map

(Kitware / Google)

Python for MATLAB Users

June 2012

97 / 201

Computer Vision in Python

Python Modules for Computer Vision

This tutorial will focus on directly using Python interfaces to large open source C++ projects:
OpenCV ITK

Several packages for computer vision are built on top of some combination of NumPy, SciPy, OpenCV, and various other modules SimpleCV http://simplecv.org/ PyVision http://sourceforge.net/projects/pyvision/ PyCVF http://pycvf.sourceforge.net/ PyCam http://code.google.com/p/pycam/ NDVision https://launchpad.net/ndvision

(Kitware / Google)

Python for MATLAB Users

June 2012

99 / 201

Loading and Saving Images with PIL

The Python Imaging Library (PIL) has many image manipulation functions. PIL is most useful for loading and saving images in a large variety of standard image le formats. PIL uses its own image class, but it is easily converted to/from NumPy arrays.
import Image import numpy as np def load_image ( file_name ): """ Read an image file and return a NumPy array . """ return np . array ( Image . open ( file_name )) def save_image ( data , file_name ): """ Save a NumPy array in an image file . """ return Image . fromarray ( data ). save ( file_name )

(Kitware / Google)

Python for MATLAB Users

June 2012

100 / 201

What is OpenCV?

C/C++ library with Python wrappers Open source Very active community Implementations of many computer vision algorithms

(Kitware / Google)

Python for MATLAB Users

June 2012

101 / 201

OpenCV Functionality Overview

Graphical User Interface(GUI) - very easy procedural interface Input/Output - load/write images and movies in a variety of formats Image Processing - ltering,edges,histograms, etc. Machine Learning - SVMs, Decision Trees, Features Object Detection and MORE! (http://docs.opencv.org/)

(Kitware / Google)

Python for MATLAB Users

June 2012

102 / 201

OpenCV By Example Importing

Two versions of wrappers: cv and cv2 cv2 contains cv cv2 is preferred because it natively uses numpy Examples/opencv rand.py
3 4 import cv2 import numpy as np

(Kitware / Google)

Python for MATLAB Users

June 2012

103 / 201

OpenCV By Example Display Random Numpy Array

Examples/opencv rand.py
1 2 3 4 5 6 7 8 9 10 # !/ usr / bin / python import cv2 import numpy as np

a = np . random . rand (256 , 256) cv2 . imshow ( Random Image , a ) cv2 . waitKey () cv2 . destroyWindow ( Random Image )

Run Examples/opencv_rand.py

(Kitware / Google)

Python for MATLAB Users

June 2012

104 / 201

OpenCV By Example Display Random Numpy Array

Generate an array of random numbers with NumPy.
7 a = np . random . rand (256 , 256)

Show the array as an image in an OpenCV GUI window. The string is both the window identier and the window title.
8 cv2 . imshow ( Random Image , a )

Wait for the user to press a key. Optionally this function can take a timeout in milliseconds.
9 cv2 . waitKey ()

Close the window and free memory. This is not strictly required in this case because it happens automatically when the script terminates.
10 cv2 . destroyWindow ( Random Image )

(Kitware / Google)

Python for MATLAB Users

June 2012

105 / 201

OpenCV By Example Image Processing

Examples/opencv hello.py
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 # !/ usr / bin / python import cv2 import sys if len ( sys . argv ) < 2: # with no arguments , use this default file filename = " / home / tutorial / Tutorial / Data / photo . jpg " elif len ( sys . argv ) == 2: filename = sys . argv [1] else : sys . exit ( " Expecting a single image file argument " ) image = cv2 . imread ( filename ) print image . shape image_small = cv2 . resize ( image , (800 , 600)) textColor = (0 , 0 , 255) # red cv2 . putText ( image_small , " Hello World !!! " , (200 , 200) , cv2 . FONT_HERSHEY_PLAIN , 3.0 , textColor , thickness =4 , linetype = cv2 . CV_AA ) cv2 . imshow ( Hello World GUI , image_small ) cv2 . waitKey () cv2 . d e s t r oy All Win dows ()

(Kitware / Google)

Python for MATLAB Users

June 2012

106 / 201

OpenCV By Example Image Processing

Run Examples/opencv_hello.py Data/photo.jpg

(Kitware / Google)

Python for MATLAB Users

June 2012

107 / 201

OpenCV By Example Image Processing

Use OpenCV to read the image le into a NumPy array
14 image = cv2 . imread ( filename )

Resize the image to 800 600 using bilinear interpolation

16 image_small = cv2 . resize ( image , (800 , 600))

Dene a tuple to represent a color in BGR color space

17 textColor = (0 , 0 , 255) # red

Draw text on the image in at location (200, 200) at a scale of 3.0

18 19 20 cv2 . putText ( image_small , " Hello World !!! " , (200 , 200) , cv2 . FONT_HERSHEY_PLAIN , 3.0 , textColor , thickness =4 , linetype = cv2 . CV_AA )

(Kitware / Google)

Python for MATLAB Users

June 2012

108 / 201

OpenCV By Example Movie Player

Examples/opencv movie.py
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 # !/ usr / bin / python import cv2 import sys if len ( sys . argv ) < 2: # with no arguments , use this default file filename = " / home / tutorial / Tutorial / Data / movie . mov " elif len ( sys . argv ) == 2: filename = sys . argv [1] else : sys . exit ( " Expecting a single video file argument " ) movie = cv2 . VideoCapture ( filename ) while True : valid_frame , image = movie . read () if not valid_frame : break cv2 . imshow ( OpenCV Video Example , image ) cv2 . waitKey (5) cv2 . d e s t r oy All Win dows ()

Run Examples/opencv_movie.py Data/movie.mov

(Kitware / Google) Python for MATLAB Users June 2012 109 / 201

OpenCV By Example Movie Player

Open the video le
14 movie = cv2 . VideoCapture ( filename )

Start an innite loop for reading and displaying video frames

15 while True :

Read one frame if available, if not exit the loop

16 17 18 valid_frame , image = movie . read () if not valid_frame : break

Display the image in a GUI window, and wait for 5 milliseconds

19 20 cv2 . imshow ( OpenCV Video Example , image ) cv2 . waitKey (5)

(Kitware / Google)

Python for MATLAB Users

June 2012

110 / 201

OpenCV By Example Feature Detection

Examples/opencv features.py
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 # !/ usr / bin / python import cv2 import sys if len ( sys . argv ) < 2: # with no arguments , use this default file filename = " / home / tutorial / Tutorial / Data / photo . jpg " elif len ( sys . argv ) == 2: filename = sys . argv [1] else : sys . exit ( " Expecting a single image file argument " ) image = cv2 . imread ( filename ) image_small = cv2 . resize ( image , ( image . shape [1] / 4 , image . shape [0] / 4)) image_gray = cv2 . cvtColor ( image_small , cv2 . cv . CV_RGB2GRAY ) features = cv2 . g o o d F e a tu r e s T o T r a c k ( image_gray , 50 , .01 , 50) features = features . reshape (( -1 , 2)) for x , y in features : cv2 . circle ( image_small , (x , y ) , 10 , (0 , 0 , 255)) cv2 . imshow ( OpenCV Features Example , image_small ) cv2 . waitKey () cv2 . d e s t r oy All Win dows ()

(Kitware / Google)

Python for MATLAB Users

June 2012

111 / 201

OpenCV By Example Feature Detection

Run Examples/opencv_features.py Data/photo.jpg

(Kitware / Google)

Python for MATLAB Users

June 2012

112 / 201

OpenCV By Example Feature Detection

Convert the color space from RGB to Gray. Feature detection requires a gray-scale image.
16 image_gray = cv2 . cvtColor ( image_small , cv2 . cv . CV_RGB2GRAY )

Run the Good Features to Track algorithm to nd 50 points. The last two arguments are a threshold and minimum distance between points.
17 features = cv2 . g o o d F e a tu r e s T o T r a ck ( image_gray , 50 , .01 , 50)

The features are a 3D array of shape (50, 1, 2). Reshape it to (50, 2). The -1 means infer from data length.
18 features = features . reshape (( -1 , 2))

At each feature location (x, y) draw a circle of radius 10 in red.

19 20 for x , y in features : cv2 . circle ( image_small , (x , y ) , 10 , (0 , 0 , 255))

(Kitware / Google)

Python for MATLAB Users

June 2012

113 / 201

Exercise 3
OpenCV

Open the le Exercises/ex3.py. Modify the code to run goodFeaturesToTrack on the movie. Then modify it to run Canny edge detection instead. Exercises/ex3.py
16 17 18 19 20 21 22 23 24 while True : valid_frame , image = movie . read () if not valid_frame : break # Insert code here to process the image before displaying it cv2 . imshow ( Exercise 3: OpenCV , image ) cv2 . waitKey (5)

(Kitware / Google)

Python for MATLAB Users

June 2012

114 / 201

Exercise 3: Answer
OpenCV

Exercises/Answers/ex3-answer1.py
19 20 21 22 23 24 image_gray = cv2 . cvtColor ( image , cv2 . cv . CV_RGB2GRAY ) features = cv2 . g o o d F e a t ur e s T o T r a c k ( image_gray , 50 , .01 , 50) if features is not None : for x , y in features . reshape (( -1 , 2)): cv2 . circle ( image , (x , y ) , 10 , (0 , 0 , 255)) cv2 . imshow ( Exercise 3: OpenCV , image )

Exercises/Answers/ex3-answer2.py
19 20 21 image_gray = cv2 . cvtColor ( image , cv2 . cv . CV_RGB2GRAY ) edges = cv2 . Canny ( image_gray , 100 , 120) cv2 . imshow ( Exercise 3: OpenCV , edges )

(Kitware / Google)

Python for MATLAB Users

June 2012

115 / 201

Simple ITK

Introduction to ITK Introduction to SimpleITK Getting SimpleITK Basic Operations Advanced Operations Building Bridges with Other Libraries

(Kitware / Google)

Python for MATLAB Users

June 2012

116 / 201

What is ITK ?

C++ Library Open Source (Apache 2.0 License) Generic Programming / C++ Templates Image Processing Image Segmentation Image Registration ITKs main foci are segmentation and registration, especially in the context of 3D images.

(Kitware / Google)

Python for MATLAB Users

June 2012

117 / 201

Funding

Funded (mostly) by the US National Library of Medicine With contributions from

National National National National National National National Institute of Dental and Craniofacial Research Science Foundation Eye Institute Institute of Neurological Disorders and Stroke Institute of Mental Health Institute on Deafness and Other Communication Disorders Cancer Institute

(Kitware / Google)

Python for MATLAB Users

June 2012

118 / 201

History

Started in 2000 Developed by Companies and Universities

GE Corporate Research Insightful Kitware UNC Chapel Hill University of Pennsylvania University of Utah

(Kitware / Google)

Python for MATLAB Users

June 2012

119 / 201

History

Recent contributions by a larger community

Harvard - Brigham and Womens Hospital University of Iowa Georgetown University INRA - France German Cancer Research Center . . . and many others . . .

(Kitware / Google)

Python for MATLAB Users

June 2012

120 / 201

Project Prole

Total lines of code: 1,886,674 Active developers: 56 (past 12 months) Total developers: 146 (full history of the project) Top 2% largest open source teams in the world Estimated cost of development: $34.5 M1

Ohloh: http://www.ohloh.net/p/itk
(Kitware / Google) Python for MATLAB Users June 2012 121 / 201

Funding

First 10 years of development = $15M Refactoring of ITKv4 in 2011 = $5M Yearly maintenance supported by NLM

(Kitware / Google)

Python for MATLAB Users

June 2012

122 / 201

Simple ITK

SimpleITK provides a template-free layer on top of ITK that is automatically wrapped for Python.

More info: http://simpleitk.org (Kitware / Google) Python for MATLAB Users June 2012 123 / 201

Installing Simple ITK

Two main options are building from source with Python wrappings Or easy install (easy install is much easier).
bash ~/ easy_install SimpleITK # It s that easy ( you may have to be root ).

Lets walk through an example of basic Simple ITK usage from Examples/sitk intro.py.

(Kitware / Google)

Python for MATLAB Users

June 2012

124 / 201

Creating Images with Simple ITK

Its very easy to create images using SimpleITK. The images will be two-dimensional or three-dimensional based on the constructor Examples/sitk intro.py
4 5 6 7 8 import SimpleITK as sitk # Create Images image = sitk . Image (256 , 256 , 256 , sitk . sitkInt16 ) twoD = sitk . Image (64 , 64 , sitk . sitkFloat32 )

(Kitware / Google)

Python for MATLAB Users

June 2012

125 / 201

Creating Images with Simple ITK

sitk is the module Image is the constructor for the Image class Height, width, depth (omit depth for 2D images) Datatype (more on this later)

(Kitware / Google)

Python for MATLAB Users

June 2012

126 / 201

What just happened?

Examples/sitk intro.py
11 12 13 print image . GetPixel (0 , 0 , 0) image . SetPixel (0 , 0 , 0 , 1) print image . GetPixel (0 , 0 , 0)

Get the voxel value at [0,0,0]? Hmm, I dont like it, so set to 1 What is the value at [0,0,0] now?

(Kitware / Google)

Python for MATLAB Users

June 2012

127 / 201

What just happened?

Examples/sitk intro.py
16 17 18 print image [0 , 0 , 0] image [0 , 0 , 0] = 10 print image [0 , 0 , 0]

Without warning, we sprinkled syntactic sugar on you! image[0,0,0] is shorthand for image.GetPixel(0,0,0) image[0,0,0] = 10 is shorthand for image.SetPixel(0,0,0,10)

(Kitware / Google)

Python for MATLAB Users

June 2012

128 / 201

SimpleITK Pixel Types

Simple ITK supports a wide array of pixel types To get a human readable assessment of the pixel type: Examples/sitk intro.py
21 print image . G e t P i x e l I D T y p e A s S t r i n g ()

(Kitware / Google)

Python for MATLAB Users

June 2012

129 / 201

SimpleITK Image Summary

Images are created using SimpleITK.Image ( w, h, d, Type ) Images can be 2- or 3-dimensional Images can describe themselves Images have simple pixel accessors

(Kitware / Google)

Python for MATLAB Users

June 2012

130 / 201

Filtering with SimpleITK

Basic operations are very simple Lets look at an image with a Gaussian subtracted from it: Examples/sitk intro.py
22 smooth = sitk . S m o o t h i n g R e c u r s i v e G a u s s i a n ( image , 2.0)

The output of SmoothingRecursiveGaussian is of type oat The input image is signed short Most SimpleITK lters with 2 inputs require the same type We need to cast in order to dierence the input image with the smoothed image

(Kitware / Google)

Python for MATLAB Users

June 2012

131 / 201

Introducing Cast

The Cast Filter works as you would expect. Examples/sitk intro.py

25 26 27 28 print " Before : " , smooth . G e t P i x e l I D T y p e A s S t r i n g () smooth = sitk . Cast ( smooth , image . GetPixelIDValue ()) print " After : " , smooth . G e t P i x e l I D T y p e A s S t r i n g () sitk . Show ( sitk . Subtract ( image , smooth ) , " DiffWithGaussian " )

(Kitware / Google)

Python for MATLAB Users

June 2012

132 / 201

Sizes and Indices

Extracting a region of interest from a SimpleITK image is easy with sitk.Extract. Examples/sitk intro.py
31 32 33 size = [64 , 64 , 0] start = [64 , 0 , 0] sitk . Show ( sitk . Extract ( image , size , start ) , " Extracted " )

(Kitware / Google)

Python for MATLAB Users

June 2012

133 / 201

More Advanced Filtering

Now that we can do basic ltering, lets look at more advanced lters How about Otsu Thresholding? Examples/sitk example.py
11 12 13 14 15 16 17 import SimpleITK as sitk image = sitk . ReadImage ( filename ) sitk . Show ( image , Slice ) image = sitk . Cast ( image , sitk . sitkFloat32 ) thresh = sitk . OtsuThreshold ( image , 0 , 230 , 15) sitk . Show ( thresh , Thresholded )

Notice that lters can also be instantiated as objects

(Kitware / Google)

Python for MATLAB Users

June 2012

134 / 201

More Advanced Filtering

We can run nearly every lter from ITK in this manner and quickly view results (as shown below). Bonus: sitk.Show allows us to view 3D Images with ImageJ

(Kitware / Google)

Python for MATLAB Users

June 2012

135 / 201

Integration with NumPy

Lets look at integrating SimpleITK with NumPy. Examples/sitk numpy.py
11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 import SimpleITK as sitk import numpy as np import pylab image = sitk . ReadImage ( filename ) a = sitk . Get Arr ayF romI mag e ( image ) h , bins = np . histogram (a , bins =255 , range =(0 , 255)) mode = 0 for i in range (1 , h . shape [0]): if h [ i ] > h [ mode ]: mode = i print " Mode of image : " + str ( mode ) # Print histogram pylab . figure () pylab . hist ( a . flatten ()) pylab . show ()

(Kitware / Google)

Python for MATLAB Users

June 2012

136 / 201

Exercise 4
SimpleITK

Open the le Exercises/ex4.py. Lets take the previous examples and build on them.
1

Use sitk.RecursiveGaussian to smooth the image, then plot the histogram with Matplotlib. Use sitk.Threshold to segment bone from the input image.

Hint: Bone has intensity greater than 210 in this image.

(Kitware / Google)

Python for MATLAB Users

June 2012

137 / 201

Exercise 4: Answer
SimpleITK

Great Job! your code should look something like this Exercises/Answers/ex4-answer.py
11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 import SimpleITK as sitk import pylab image = sitk . ReadImage ( filename ) image_array = sitk . GetA rra yFr omI mage ( image ) sitk . Show ( image ) blurred = sitk . R ecur siv eGa uss ian ( image , 4) blurred_array = sitk . Get Arr ayF rom Imag e ( blurred ) # Generate a normal and blurred histogram pylab . figure () pylab . hist ( image_array . flatten ()) pylab . figure () pylab . hist ( blurred_array . flatten ()) pylab . show () # Threshold the image properly thresh = sitk . Threshold ( image , 210 , 255) sitk . Show ( thresh )

(Kitware / Google)

Python for MATLAB Users

June 2012

138 / 201

More Advanced Filtering

Your bone segmentation should look something like this.

(Kitware / Google)

Python for MATLAB Users

June 2012

139 / 201

Credits

Material taken from the following tutorials: The SimpleITK tutorial from MICCAI 2011 (https: //github.com/SimpleITK/SimpleITK-MICCAI-2011-Tutorial). The ITKv4 next generation tutorial (https://github.com/InsightSoftwareConsortium/ ITKv4-TheNextGeneration-Tutorial).

(Kitware / Google)

Python for MATLAB Users

June 2012

140 / 201

Parallel Processing in Python

Multi-Threading in Python

Python supports multi-threading of code but has an important limitation. Only one thread can execute Python code at a time. Multi-thread can still be useful in certain cases such as
running multiple I/O-bound tasks simultaneously code that uses a lot of C-based modules (e.g. NumPy and SciPy)

To get parallel processing in Python it is often better to use multi-processing.

http://docs.python.org/library/threading.html (Kitware / Google) Python for MATLAB Users June 2012 142 / 201

Multi-Processing in Python
The multiprocessing module provides convenient ways to divide Python execution over multiple processes. Create a pool of worker processes and divide up the work. Examples/multiproc.py
1 2 3 4 5 6 7 8 9 10 from m ul tiprocessing import Pool def f ( x ): return x * x if __name__ == __main__ : pool = Pool ( processes =4) result = pool . apply_async (f , [10]) print result . get ( timeout =1) print pool . map (f , range (10))

# # # #

start 4 worker processes evaluate " f (10)" asynchronously prints "100" unless your computer is slow prints "[0 , 1 , 4 ,... , 81]"

http://docs.python.org/library/multiprocessing.html (Kitware / Google) Python for MATLAB Users June 2012 143 / 201

Modern GPGPU
GPUs have been evolving from dedicating graphics rendering hardware, to powerful general computing devices:

Modern GPU Architecture (Nvidia Kepler)

http://blogs.nvidia.com/tag/kepler/index.html (Kitware / Google) Python for MATLAB Users June 2012 144 / 201

Modern GPGPU
Massive amount of processing cores (3072 CUDA cores in Nvidia GeForce GTX 690 single card) Each core has its own dedicated register le Each cluster of cores share L1 cache Clusters of cores share L2 cache Huge memory throughput (512bit, 384GB/sec for Nvidia GeForce 690) Atomic operations 64-bit precision Error-correcting Code (ECC) memory to meet HPC requirement GPU Vs CPU on HPC performance
Most CV developers: 10 to 500x speed-up depending on the application and implementation Intel: GPUs are ONLY up to 14 times faster than CPU.
http://blogs.nvidia.com/tag/kepler/index.html (Kitware / Google) Python for MATLAB Users June 2012 145 / 201

GPGPU Programming Model: Shading Language

Map parallel computing task to openGL rendering pipeline Pack input data as vertices and texture, operate these data from vertex shader and fragment shader, results are output to frame buer Pros:
enable the GPU for general use for the rst time
Shading Language (Vertex & Fragment Shader)

Cons:
single precision no ow control and branching very limited kernel size lack of cross-kernel synchronization

(Kitware / Google)

Python for MATLAB Users

June 2012

146 / 201

GPGPU Programming Model: CUDA and OpenCL

Enable C/C++ like direct parallel thread programming Enable cross-thread synchronization Enable direct memory access Enable atomic operations Eectively remove the kernel size limit Enable full support for ow control and branching GPUs become truly general purpose
CUDA Programming Model (OpenCL has very similar model)
http://ixbtlabs.com/articles3/video/cuda-1-p5.html (Kitware / Google) Python for MATLAB Users June 2012 147 / 201

GPGPU in Computer Vision

Many hardwired graphics rendering resources in GPU, including texture engine, 3D transformation, make GPU a natural t for computer vision and computational photography processing. OpenVIDIA: Parallel GPU Computing Vision Library developed by Nvidia OpenCV: more and more GPU Modules Nvidia CUDA Zone: An Image and Computer Vision GPU library, including segmentation, feature processing, stereo imaging, machine learning and data processing

(Kitware / Google)

Python for MATLAB Users

June 2012

148 / 201

Why use GPGPU in Python?

GPUs are everything that scripting languages are not

Highly parallel Very architecture-sensitive Built for maximum performance

GPU and Python can be complement to each other

Use Python script as the brain, organizing processing module and data connection Use GPU as muscles, executing intensive, dirty inner loops

Play to the strengths of each programming environment Python + CUDA: PyCUDA Python + OpenCL: PyOpenCL

http://mathema.tician.de/software/pycuda (Kitware / Google) Python for MATLAB Users June 2012 149 / 201

Native CUDA work ow

(Kitware / Google)

Python for MATLAB Users

June 2012

150 / 201

PyCUDA work ow

http://mathema.tician.de/software/pycuda (Kitware / Google) Python for MATLAB Users June 2012 151 / 201

PyCUDA Example

import pycuda . driver as cuda import pycuda . autoinit , pycuda . compiler import numpy

# import PyCUDA module # initialize the CUDA hardware

a = numpy . random . randn (4 ,4). astype ( numpy . float32 ) # create a 4 x4 array random numbers a_gpu = cuda . mem_alloc ( a . nbytes ) # alloc the same sized array in GPU cuda . memcpy_htod ( a_gpu , a ) # copy the data from host to device mod = pycuda . compiler . SourceModule ( """ # specify a CUDA kernel function __global__ void twice ( float * a ) { int idx = threadIdx . x + threadIdx . y *4; a [ idx ] *= 2; } """ ) func = mod . get_function ( " twice " ) # get the handle of CUDA kernel func ( a_gpu , block =(4 ,4 ,1)) # run the CUDA function with a configuration a_doubled = numpy . empty_like ( a ) cuda . memcpy_dtoh ( a_doubled , a_gpu ) print a_double print a # create an empty buffer on host # copy result from device to host

http://mathema.tician.de/software/pycuda (Kitware / Google) Python for MATLAB Users June 2012 152 / 201

gpuarray: Simple Linear Algebra

Meant to look and feel just like numpy +, -, *, /, fill, sin, exp, rand, basic indexing, norm, inner product, ... Mixed types (int32 + float32 = float64) print gpuarray for debugging Allows access to raw bits

http://mathema.tician.de/software/pycuda (Kitware / Google) Python for MATLAB Users June 2012 153 / 201

gpuarray: Element-wise expressions

An example of element-wise operation in PyCUDA:

from pycuda . curandom import rand as curand a_gpu = curand ((50 ,)) b_gpu = curand ((50 ,)) # import CUDA random number module # create a 1 - d array with random number

from pycuda . elementwise import El eme ntwi seK ern el

# import El eme ntw ise Kern el module

# specify the detail of element - wise operation lin_comb = Ele men twi seK ern el ( " float a , float *x , float b , float *y , float * z " , "z[i] = a*x[i] + b*y[i]") c_gpu = gpuarray . empty_like ( a_gpu ) # create a GPU array of same size lin_comb (5 , a_gpu , 6 , b_gpu , c_gpu ) # run the E lem e ntw ise Ker nel function assert la . norm (( c_gpu - (5* a_gpu +6* b_gpu )). get ()) < 1e -5 print a_gpu print b_gpu print c_gpu

http://mathema.tician.de/software/pycuda (Kitware / Google) Python for MATLAB Users June 2012 154 / 201

gpuarray: Reduction made easy

Example: A scalar product calculation
from pycuda . reduction import ReductionKernel # import ReductionKernel module

# specify the detail of the reduction operation dot = Re ductionKernel ( dtype_out = numpy . float32 , neutral = " 0 " , reduce_expr = " a + b " , map_expr = " x [ i ]* y [ i ] " , arguments = " const float *x , const float * y " ) from pycuda . curandom import rand as curand x = curand ((1000*1000) , dtype = numpy . float32 ) y = curand ((1000*1000) , dtype = numpy . float32 ) x_dot_y = dot (x , y ). get () x_dot_y_cpu = numpy . dot ( x . get () , y . get ()) print print print print x y x_dot_y x_dot_y_cpu

http://mathema.tician.de/software/pycuda (Kitware / Google) Python for MATLAB Users June 2012 155 / 201

PyCUDA: Vital Information

http://mathema.tician.de/software/pycuda Complete documentation MIT License (no warranty, free for all use) Requires: numpy, Python 2.4+ (Win/OS X/Linux) Support via mailing list

http://mathema.tician.de/software/pycuda (Kitware / Google) Python for MATLAB Users June 2012 156 / 201

Introducing... OpenCL and PyOpenCL

OpenCL: the Open Computing Language

Vendor-independence Single abstraction works well for GPUs, CPUs A JIT C compiler baked into a library Intels future integrated GPUs will be supporting (more ubiquitous, available in cloud)

PyOpenCL: the PyCUDA for OpenCL

Complete, mature API wrapper Has: Arrays, elemet-wise operations, ... http://mathema.tician.de/software/pyopencl

http://mathema.tician.de/software/pyopencl (Kitware / Google) Python for MATLAB Users June 2012 157 / 201

OpenCL: Same avor, dierent recipe

import pyopencl as cl , numpy

# import PyOpenCL module # create a buffer on host

a = numpy . random . rand (50000). astype ( numpy . float32 ) ctx = cl . c r ea t e _ s o m e _ co n t e x t () queue = cl . CommandQueue ( ctx )

# create an OpenCL context

# create a buffer on device and copy the data in host buffer to device buffer a_buf = cl . Buffer ( ctx , cl . mem_flags . READ_WRITE , size = a . nbytes ) cl . e n q u e u e _ w r i t e _ b u f f e r ( queue , a_buf , a ) # specify an OpenCL kernel function prg = cl . Program ( ctx , """ __kernel void twice ( __global float * a ) { int gid = get_global_id (0); a [ gid ] *= 2; } """ ). build () # run an OpenCL kernel function prg . twice ( queue , a . shape , None , a_buf ). wait ()

http://mathema.tician.de/software/pyopencl (Kitware / Google) Python for MATLAB Users June 2012 158 / 201

Metaprogramming

GPU code does not need to be a compile-time constant Code is data - it wants to be reasoned about at run time Automated tuning Data types Specialized code for given problem Constants faster than variables Loop unrolling

http://mathema.tician.de/software/pycuda (Kitware / Google) Python for MATLAB Users June 2012 159 / 201

Showcase...

(Kitware / Google)

Python for MATLAB Users

June 2012

160 / 201

Interfacing Python with Other Languages

Wrapping C/C++ code in Python

There are several options for wrapping C/C++ code in Python Calling the Python C API Manually (not for the faint of heart) Interfacing with shared libraries using ctypes Using Cython to generate C API calls (good for optimizing critical pieces of code) Using SWIG to generate C API calls (good for wrapping larger libraries) Using Boost.Python to bridge between Python and C++

(Kitware / Google)

Python for MATLAB Users

June 2012

162 / 201

Working with the Python C API

Calling the C API directly allows you to get the most performance possible and have the greatest control of execution. This is also one of the least easily maintainable methods. For more information, see http://docs.python.org/extending/extending.html. There is a simple example at Examples/c-api-demonstration/cdemo.c

(Kitware / Google)

Python for MATLAB Users

June 2012

163 / 201

A Note on the ctypes Module

ctypes is a built-in module that allows calling functions from shared libraries. A more exhaustive explanation of what ctypes can do can be found here: http://docs.python.org/library/ctypes.html. For example, we can easily call the standard libc time function with the following snippet of code:
7 8 9 10 11 12 13 14 15 16 17 from ctypes import cdll , CDLL # The name of our library library_name = libc . so .6 # Load the library as a module cdll . LoadLibrary ( library_name ) libc = CDLL ( library_name ) # Call the standard time function and print the results print libc . time ( None )

This can be very useful, but doesnt feel as pythonic as some of the other approaches.
(Kitware / Google) Python for MATLAB Users June 2012 164 / 201

Cython

Cython is a tool for compiling Python-like code into C to be compiled into python extension modules. There are sets of annotations that can further optimize code through static typing, removal of bounds checking, etc. An example of a pyx le and generated C le can be found in Examples/cython Lets go through a simple example.

More Info: http://www.cython.org/ (Kitware / Google) Python for MATLAB Users June 2012 165 / 201

Cython

This simple code compiled with Cython is 35% faster.

0 1 2 3 4 5 6 7 8 def f ( x ): return x **2 - x def integrate_f (a , b , N ): s = 0 dx = (b - a )/ N for i in range ( N ): s += f ( a + i * dx ) return s * dx

By adding type declarations we can go a great deal faster, even with this simple example.

(Kitware / Google)

Python for MATLAB Users

June 2012

166 / 201

Cython

This code compiled with Cython is 400% faster, but we lose compatibility with Python through the calls to cdef.
0 1 2 3 4 5 6 7 8 9 10 def f ( double x ): return x **2 - x def integrate_f ( double a , double b , int N ): cdef int i cdef double s , dx s = 0 dx = (b - a )/ N for i in range ( N ): s += f ( a + i * dx ) return s * dx

We can do even better by typing the function call.

(Kitware / Google)

Python for MATLAB Users

June 2012

167 / 201

Cython

This code compiled with Cython is 15,000% faster, but the calls are no able to raise Python errors.
0 1 2 3 4 5 6 7 8 9 10 cdef double f ( double x ) except ? -2: return x **2 - x def integrate_f ( double a , double b , int N ): cdef int i cdef double s , dx s = 0 dx = (b - a )/ N for i in range ( N ): s += f ( a + i * dx ) return s * dx

(Kitware / Google)

Python for MATLAB Users

June 2012

168 / 201

Cython

Lets take a look at that timing on our virtual machines using

In [1]: import simple In [2]: import simplec In [3]: import simple_opt In [4]: import simple_opt_full In [5]: % timeit for x in range (1 ,1001): simple . integrate_f (3 , 5 , x ) 10 loops , best of 3: 137 ms per loop In [6]: % timeit for x in range (1 ,1001): simplec . integrate_f (3 , 5 , x ) 10 loops , best of 3: 79.8 ms per loop In [7]: % timeit for x in range (1 ,1001): simple_opt . integrate_f (3 , 5 , x ) 10 loops , best of 3: 55.6 ms per loop In [8]: % timeit for x in range (1 ,1001): simple_opt_full . integrate_f (3 , 5 , x ) 1000 loops , best of 3: 1.85 ms per loop

(Kitware / Google)

Python for MATLAB Users

June 2012

169 / 201

Cython

Here is an example of how to compile your cython into a loadable module (taken from http://docs.cython.org/src/reference/compilation.html).
0 1 2 $ cython yourmod . pyx $ gcc - shared - pthread - fPIC - fwrapv - O2 - Wall -fno - strict - aliasing \ -I / usr / include / python2 .7 -o yourmod . so yourmod . c

This should give you an idea of the possibilities for rapid optimization with Cython. See http://cython.org/ for more information.

(Kitware / Google)

Python for MATLAB Users

June 2012

170 / 201

SWIG

SWIG is a great option for wrapping existing C/C++ code for many dierent languages, including Python An example of swig usage can be found in Examples/swig SWIG is generally used for large projects that need semi-automated wrapping for multiple languages (SimpleITK uses this for its wrapping).

More Info: http://www.swig.org/ (Kitware / Google) Python for MATLAB Users June 2012 171 / 201

SWIG
Here is an example of a C le we want to automatically wrap with SWIG
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 # include < time .h > double My_variable = 3.0; int fact ( int n ) { if ( n <= 1) return 1; else return n * fact (n -1); } int my_mod ( int x , int y ) { return ( x % y ); } char * get_time () { time_t ltime ; time (& ltime ); return ctime (& ltime ); }

(Kitware / Google)

Python for MATLAB Users

June 2012

172 / 201

SWIG

In SWIG we dene the interface or .i les

2 3 4 5 6 7 8 9 10 11 12 13 14 % module example %{ /* Put header files here or function declarations like below */ extern double My_variable ; extern int fact ( int n ); extern int my_mod ( int x , int y ); extern char * get_time (); %} extern extern extern extern double My_variable ; int fact ( int n ); int my_mod ( int x , int y ); char * get_time ();

(Kitware / Google)

Python for MATLAB Users

June 2012

173 / 201

SWIG

To generate our example wrapping and build example.c itself, we run the following commands.
3 4 5 swig - python example . i gcc -c example . c example_wrap . c -I / usr / include / python2 .7 ld - shared example . o example_wrap . o -o _example . so

Swig generates a shared library and python wrapper for importing it as shown below.
In [9]: import example In [10]: example . fact (10) Out[10]: 3628800

(Kitware / Google)

Python for MATLAB Users

June 2012

174 / 201

SWIG

This pattern can be used to wrap existing C and C++ libraries in a maintainable way. There is also support for converting between STL types and Python types.

(Kitware / Google)

Python for MATLAB Users

June 2012

175 / 201

Boost.Python
Boost.Python is an alternative to SWIG.
Only for making Python bindings (not multi-language) Designed to make wrapping of C++ easy. Handles C++ classes, virtual functions, inheritance, etc. Only involves writing C++ code (no interface les)

To wrap the previous example code

# include < boost / python . hpp > B O O S T _ P Y T H O N _ M OD U L E ( example ) { using namespace boost :: python ; scope (). attr ( " My_variable " ) = My_variable ; def ( " fact " , fact ); def ( " my_mod " , my_mod ); def ( " get_time " , get_time ); }

More Info: http://www.boost.org/doc/libs/1_49_0/libs/python/doc/ (Kitware / Google) Python for MATLAB Users June 2012 176 / 201

Interfacing with Matlab

Interfacing with Matlab

Any transition to Python from Matlab will require a means to utilize your previous Matlab code base. There are several methods for doing this: 1 Converting Matlab Code to Python
Open-Source Matlab-to-Python Compiler (OMPC) (ompc.juricap.com):
Converts existing Matlab functions to Python, does not require Matlab license, but only works for simple functions
2

File Interchange with Matlab

Load/Save .mats in Python:
Scipy.io.< loadmat /savemat > load/save .mats, does not require Matlab license, but may need to manipulate format after loading

Calling Legacy MATLAB Code from Python

Mlabwrap(mlabwrap.sourceforge.net):
Calls Matlab functions in manner similar to Matlab, does require Matlab license, and slower than Matlab

(Kitware / Google)

Python for MATLAB Users

June 2012

178 / 201

OMPC Example
Convert a Simple Matlab function that plots random numbers to Python Examples/plot rand nums ex1.m Convert .m to .pym
In [11]: import ompc In [12]: import p lo t_ r an d _n um s _e x 1 In [13]: p l o t _r a nd _n u ms _ ex 1 (100) 1 2 3 4 5 6 function p l ot _ ra nd _ nu m s_ ex 1 ( maxX ) % Function to plot a vector of random numbers x = 1: maxX ; randx = rand (1 , maxX ); plot (x , randx )

Examples/plot rand nums ex1.pym

1 2 3 4 5 6 7 8 9 10 11 12 http://ompc.juricap.com/download (Kitware / Google) Python for MATLAB Users June 2012 179 / 201 # This file was automatically translated # by OMPC ( http :// ompc . juricap . com ) from ompc import * @mfunction ( " " ) def p lo t _r an d _n u ms _e x 1 ( maxX = None ): # Function to plot a vector of random numbers x = mslice [1: maxX ] randx = rand (1 , maxX ) plot (x , randx )

Load/Save .mat Examples

Create Structure in Matlab, save as .mat, and load into Python
MATLAB >> >> >> xx xx . range = 1:11; xx . name = vars1 ; xx . cellvalues = {[15 16] , var1 ,[17 18] , var2 } = range : [1 x11 double ] name : vars1 cellvalues : {1 x4 cell } >> save ( xx . mat , xx )

Python > import scipy . io as sio > xx = sio . loadmat ( xx . mat ) > print xx { xx : array ([[ ([[1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , 11]] , [ u vars1 ] , [[ array ([[15 , 16]] , dtype = uint8 ) , array ([ u var1 ] , dtype = < U4 ) , array ([[17 , 18]] , dtype = uint8 ) , array ([ u var2 ] , dtype = < U4 )]])]] , dtype =[( range , | O4 ) , ( name , | O4 ) , ( cellvalues , | O4 )]) , __version__ : 1.0 , __header__ : MATLAB 5.0 MAT - file , Platform : PCWIN64 , Created on : Tue May 29 12:53:43 2012 , __globals__ : []}

(Kitware / Google)

Python for MATLAB Users

June 2012

180 / 201

Mlabwrap Example
Call Matlabs Surf() function from Python
1 2 3 4 5 6 7 8 9 10 11 # !/ usr / bin / python # Copyright 2011 , Kitware , Inc . # http :// mlabwrap . sourceforge . net / Examples for calling matlab functions from Python from mlabwrap import mlab ; from numpy import * def main (): Calling Matlab s surf () command xx = arange ( -2* pi , 2* pi , 0.2) # Define range of inputs mlab . surf ( subtract . outer ( sin ( xx ) , cos ( xx ))) # call matlab s surf () function

Call Matlabs Singular Value Decomposition (svd) function

13 14 Calling Matlab s Singular Value Decomposition svd () command U , S , V = mlab . svd ([[1 ,2] ,[1 ,3]] , nout =3)

More Info: http://mlabwrap.sourceforge.net/ (Kitware / Google) Python for MATLAB Users June 2012 181 / 201

Wrap-Up

Rae Prize 1

Book: A Primer on Scientic Programming with Python

(Kitware / Google) Python for MATLAB Users June 2012 183 / 201

Rae Prize 2

ASUS Transformer Pad TF300

(Kitware / Google)

Python for MATLAB Users

June 2012

184 / 201

Send Us Feedback

Please send us feedback, either positive or negative.

Thank You!

matt.leotta

amitha.perera patrick.reynolds eran.swears

yongzhao

varun

@kitware.com

@google.com

(Kitware / Google)

Python for MATLAB Users

June 2012

185 / 201

Appendix

Native Installation of Scientic Python

Native Installation of Scientic Python

The virtual machine provides a nice way to try out software without worrying about installation. Eventually youll want to install these toolkit in your native OS. The following slides provide basic installation instructions for
Windows Mac OS X Linux

(Kitware / Google)

Python for MATLAB Users

June 2012

188 / 201

Installing on Windows

It is dicult to install individual toolkits on Windows. Packages installed independently often do not work together because
package versions may be incompatible with each other packages are compiled with dierent compilers.

Lucky there is a project called Python(x,y) which installs almost everything you need for scientic Python and congures it to work together.

(Kitware / Google)

Python for MATLAB Users

June 2012

189 / 201

Installing on Windows Python(x,y)

1 2

Download the package installer (currently Python(x,y)-2.7.2.2.exe) Run the installer executable and accept the license agreement

More info: http://code.google.com/p/pythonxy/ (Kitware / Google) Python for MATLAB Users June 2012 190 / 201

Installing on Windows Python(x,y)

Select the components to install

More info: http://code.google.com/p/pythonxy/ (Kitware / Google) Python for MATLAB Users June 2012 191 / 201

Installing on Windows Python(x,y)

4 5 6

Select the installation location Select the Start Menu folder Press the Install button.

More info: http://code.google.com/p/pythonxy/ (Kitware / Google) Python for MATLAB Users June 2012 192 / 201

Installing on Windows Python(x,y)

Wait for everything to install.

More info: http://code.google.com/p/pythonxy/ (Kitware / Google) Python for MATLAB Users June 2012 193 / 201

Installing on Windows Python(x,y)

Press the Finish button and you are done!

More info: http://code.google.com/p/pythonxy/ (Kitware / Google) Python for MATLAB Users June 2012 194 / 201

Installing on OS X

This should get you setup with all the tools shown today in as clean of a way as possible.

(Kitware / Google)

Python for MATLAB Users

June 2012

195 / 201

Installing on OS X

The good news:

OS X ships with Python installed Homebrew is a great package manager for OS X (http://mxcl.github.com/homebrew/).

The bad news:

We need to install xcode Building everything can take a while

(Kitware / Google)

Python for MATLAB Users

June 2012

196 / 201

Installing on OS X

First, we assume you have XCode installed. Next, install homebrew using the instructions here: https://github.com/mxcl/homebrew/wiki/installation. Now that we have homebrew, we will install our binary dependencies
brew brew brew brew install install install install gfortran # for scipy pyqt pyside opencv

Now were ready to setup the python stu in a virtual environment.

(Kitware / Google)

Python for MATLAB Users

June 2012

197 / 201

Installing on OS X

What is a virtual environment? Python virtual environments let you keep all of your dependencies in one place. (through environment variable magic) To install virtualenv, use the instruction here: http://pypi.python.org/pypi/virtualenv Now create a virtual environment like so:
virtualenv down_with_matlab source down_with_matlab / bin / activate

Now we are ready to install our python packages and applications

(Kitware / Google)

Python for MATLAB Users

June 2012

198 / 201

Installing on OS X

The order and distinction between pip and easy install is important.
easy_install readline pip install sphinx pip install spyder pip install ipython pip install tornado easy_install zmq pip install numpy pip install scipy pip install matplotlib pip install sympy easy_install SimpleITK

Now you can use spyder, ipython, and all of these libraries as long as the virtual environment is activated. Bonus: None of this is cluttering up your systems site-packages.

(Kitware / Google)

Python for MATLAB Users

June 2012

199 / 201

Installing on Linux

The same virtualenv, pip, and easy install tools used for Mac OS X can be used for Linux, but there is an easier way. Most linux distributions have packages for scientic Python tools that make them easy to install.
Package are (usually) pre-built, so you dont have to wait for them to compile. Dependencies are automatically installed. If you install Spyder, you automatically get Python, NumPy, SciPy, etc.

A few toolkits may not be packaged and will still require pip or easy install.

(Kitware / Google)

Python for MATLAB Users

June 2012

200 / 201

Installing on Linux - Ubuntu

In Ubuntu, you can use the Ubuntu Software Center and install packages. Equivalently, from the command line you can run
sudo apt - get install spyder

to select

Installing Spyder provides most of what is needed. Other useful packages to install are
sudo apt - get install ipython - notebook ipython - qtconsole python - opencv imagej sudo easy_install SimpleItk

The apt-get commands will also work on other Debian-based OSes (like Debian and Mint).

(Kitware / Google)

Python for MATLAB Users

June 2012

201 / 201