# Python tutorial @ BIG (EPFL) ## A short introduction to Python for Image Analysis and Deep Learning *Lilian Besson*
BIG, EPFL, July 2016 --- ## Introduction This short tutorial will get you started with Python 3. We will try to discover together what Daniel asked me yesterday. --- ## 1. Installing Python 3 > *Try to do this on your laptop, during the tutorial* 1. Download [Anaconda (Python 3.5)](https://repo.continuum.io/archive/Anaconda3-4.1.0-MacOSX-x86_64.pkg) from [continuum.io/downloads](https://www.continuum.io/downloads) (~ 346 Mo) 2. Install it: double-click the downloaded `.pkg` file and follow the instructions 3. Check that Python (`python3`) has been installed: ```bash $ python3 [it should work] ``` --- ## 2. Basic introduction to Python - **Not covered today** - Start with [introtopython.org](http://introtopython.org/) - More in-depth tutorial: [scipy-lectures.org](http://www.scipy-lectures.org/) (very good quality) - Example: [Hello World!](http://introtopython.org/hello_world.html) ```python >>> print("Hello Python world!") Hello Python world! ``` --- ## 3. Using the Spyder IDE - The Spyder IDE [is shipped](https://pythonhosted.org/spyder/installation.html#installing-on-macos-x) with Anaconda - Gives a nice MATLAB-like interface: advanced editing, interactive testing, debugging and introspection features - A numerical computing environment thanks to the support of: `IPython` (enhanced interactive Python interpreter) and core Python libraries: `NumPy` (linear algebra), `SciPy` (signal and image processing) or `matplotlib` (interactive 2D/3D plotting) - Easy to debug: add breakpoint, previous/next buttons etc - → It's *Demo time!* - Other good IDE : the [Jupyter notebook](https://jupyter.org/) [jupyter.org](https://jupyter.org/) --- ## 4. Importing the main libraries - They are all shipped with Anaconda! - [NumPy](http://www.numpy.org/): `import numpy as np` - [Scipy](http://www.scipy.org/): `import scipy` - [MatPlotLib](http://matplotlib.org/): `import matplotlib.pyplot as plt` --- ## 4.1. First example: ```python t = np.linspace(0, 2 * np.pi, 400) x = np.cos(2*t) y = np.cos(3*t) plt.figure() plt.plot(x, y, 'r+-') plt.show() ``` [![Figure #1](figures/figure_1.png)](figures/figure_1.png) --- ## 4.1. Second example: ```python from scipy.special import gamma x = np.linspace(0.1, 3, 400) y = gamma(x) plt.figure() plt.plot(x, y) plt.title("The function $\Gamma(x)$ on $[0.1, 3]$") plt.show() ``` [![Figure #2](figures/figure_2.png)](figures/figure_2.png) --- ## 5. Reading data, images etc with `scipy` or `scikit-image` - They are all shipped with Anaconda! - `scipy.ndimage` implements a lot of image processing functions, mostly for n-dimensional images. Cf. the tutorial http://www.scipy-lectures.org/advanced/image_processing/index.html - And `scikit-image` ([scikit-image.org](http://scikit-image.org)) adds functions specific to 2D/3D images, and more. Cf. the tutorial http://www.scipy-lectures.org/packages/scikit-image/index.html#scikit-image - For 3D plotting, use [Mayavi](http://www.scipy-lectures.org/packages/3d_plotting/index.html#mayavi-label) --- ## 5.1. Example: reading an image ```python from scipy import ndimage # module for n-d images import matplotlib.pyplot as plt # module for plotting from scipy import misc # some toy data are in this module face = misc.face(gray=True) # Or face = plt.imread('face.png') # Or from skimage.io import imread # import a function face = imread('face.jpg') print(face[0, 0]) # first pixel: 114 # display the image plt.imshow(face, cmap='gray') plt.show() ``` [![Figure #3](figures/figure_3.png)](figures/figure_3.png) --- ## 5.2. Example: more on images ```python lx, ly = face.shape # cropping, by slicing the ndarray (matrix) crop_face = face[lx / 4: - lx / 4, ly / 4: - ly / 4] # up <-> down flip flip_ud_face = np.flipud(face) # rotation rotate_face = ndimage.rotate(face, 45) rotate_face_noreshape = ndimage.rotate(face, 45, reshape=False) plt.figure() plt.subplot(2, 3, 1) plt.imshow(face, cmap='gray') plt.subplot(2, 3, 2) plt.imshow(crop_face, cmap='gray') # etc... ``` [![Figure #4](figures/figure_4.png)](figures/figure_4.png) --- ## 6. Machine Learning in Python with `scikit-learn` - Shipped with Anaconda - Importing [scikit-learn](http://scikit-learn.org/): `import sklearn as sk`, `from sklearn import XXX` - Documentation on [scikit-learn.org](http://scikit-learn.org/stable/tutorial/index.html) - Lots of "not-deep" machine learning algorithm, easy to use - Lots of examples --- ## 7. Deep Learning in Python with `caffe`, `lasagne` or `tensorflow` - **I don't do deep learning myself!** So I don't know which library is the best... - **NOT shipped with Anaconda !** - `caffe`: Python interface to a C++ engine, by Berkeley's Vision lab, [caffe.berkeleyvision.org](http://caffe.berkeleyvision.org/), [example](https://nbviewer.jupyter.org/github/BVLC/caffe/blob/master/examples/01-learning-lenet.ipynb#2.-Creating-the-net) - `lasagne`: C and Python, built on top of `theano`, by Yoshua Bengio's lab (Montreal), [lasagne.readthedocs.org](http://lasagne.readthedocs.org/), [example](https://github.com/Lasagne/Lasagne#example) - `tensorflow`: Python interface to a C++ engine, by Google, [tensorflow.org](http://www.tensorflow.org/), [example](https://github.com/pkmital/tensorflow_tutorials#tensorflow-tutorials). See also: [tflearn.org](http://tflearn.org/#quick-overview) - Also interesting: [keras.io](http://keras.io/), using either Theano or TensorFlow, pure Python, [example](http://keras.io/#getting-started-30-seconds-to-keras) --- ## Questions ? > Please ask if any! --- ## References for Python 3 and basic tools - `Python 3` documentation: [docs.python.org/3](https://docs.python.org/3/) - [introtopython.org](http://www.introtopython.org/) for a small introduction to Python syntax and concepts - `Spyder` documentation: [pythonhosted.org/spyder](https://pythonhosted.org/spyder/) - `IPython` tutorial: [ipython.readthedocs.io](https://ipython.readthedocs.io/en/stable/interactive/tutorial.html) --- ## References for libraries (1/3) - `NumPy` documentation: [docs.scipy.org/doc/numpy/reference](http://docs.scipy.org/doc/numpy/reference/) - `SciPy` documentation: [docs.scipy.org/doc/scipy/reference](http://docs.scipy.org/doc/scipy/reference/) - `SciPy` for image manipulation: [www.scipy-lectures.org/advanced/image_processing](http://www.scipy-lectures.org/advanced/image_processing) - `MatPlotLib` documentation: [matplotlib.org/contents.html](http://matplotlib.org/contents.html) - `MatPlotLib` tutorial: [www.labri.fr/perso/nrougier/teaching/matplotlib](http://www.labri.fr/perso/nrougier/teaching/matplotlib/) --- ## References for libraries (2/3) - `scikit-learn` tutorial: [scikit-learn.org/stable/tutorial/index.html](http://scikit-learn.org/stable/tutorial/index.html) - `scikit-image` tutorial: [scikit-image.org/docs/stable/overview.html](http://scikit-image.org/docs/stable/overview.html) - Also on [scipy-lectures.org](http://www.scipy-lectures.org/): [www.scipy-lectures.org/packages/scikit-image](http://www.scipy-lectures.org/packages/scikit-image) --- ## References for libraries (3/3) - `theano` documentation: [deeplearning.net/software/theano](http://deeplearning.net/software/theano) - `lasagne` documentation: [lasagne.readthedocs.org](http://lasagne.readthedocs.org/) - `tensorflow` documentation: [www.tensorflow.org/versions/r0.9/get_started/index.html](https://www.tensorflow.org/versions/r0.9/get_started/index.html) - `tflearn` tutorial: [tflearn.org/#quick-overview](http://tflearn.org/#quick-overview) - `keras` tutorial: [keras.io/#getting-started-30-seconds-to-keras](http://keras.io/#getting-started-30-seconds-to-keras) --- ## Questions ? > Please ask if any!