SparseEdges tests¶

A series of tests for the SparseEdges package¶

Table of content

What is the SparseEdges package?
Installing
testing one step

In [1]:

%load_ext autoreload
%autoreload 2

from __future__ import division, print_function
%matplotlib inline
import matplotlib.pyplot as plt

import numpy as np
np.set_printoptions(precision=4)#, suppress=True)

In [2]:

%cd -q ../probe

In [3]:

from NeuroTools.parameters import ParameterSet
from SLIP import Image
from LogGabor import LogGabor
from SparseEdges import SparseEdges

In [4]:

#! defining framework
#!-------------------
mp = SparseEdges('https://raw.githubusercontent.com/bicv/SparseEdges/master/default_param.py')
mp.pe.N = 4
mp.pe.MP_alpha=1.
mp.pe.do_whitening = False

In [5]:

!ls -l  ../probe/cache_dir/experiment_test_*

-rw-r--r--  1 laurentperrinet  staff   6272 Jun 15 16:41 ../probe/cache_dir/experiment_test_MP.npy
-rw-r--r--  1 laurentperrinet  staff  12416 Jun 15 18:44 ../probe/cache_dir/experiment_test_nowhitening.npy
-rw-r--r--  1 laurentperrinet  staff  24704 Jun 15 18:04 ../probe/cache_dir/experiment_test_whitening.npy
-rw-r--r--  1 laurentperrinet  staff   4224 Jun 15 18:04 ../probe/cache_dir/experiment_test_whitening_MSE.npy

Defining units, axis and conventions¶

cardinal points in X, Y (clockwise): E (=right=3 o'clock=horizontal), ESE, S, WSW, W, WNW, N, ENE :

In [6]:

X = np.array([0, 1./4, 1./2, 1./4, 0, -1./4, -1./2])#, -1./4])\n",
Y = np.array([1./2, np.sqrt(3)/4, 0, -np.sqrt(3)/4, -1./2, -np.sqrt(3)/4, 0.])#, np.sqrt(3)/
Theta = np.arctan2(-X, Y)  # perpendicular to circle
Theta = ((Theta + np.pi/2  - np.pi/mp.pe.N_phi/2 ) % (np.pi)) - np.pi/2  + np.pi/mp.pe.N_phi/2
print(['%0.1f' % (angle*180/np.pi) for angle in Theta])

['-0.0', '-30.0', '90.0', '30.0', '-0.0', '-30.0', '90.0']

We set the [0,0] index of the array in the upper left left corner of the axes.

The X, Y image axis are defined in the matrix style (origin on the upper left; X going vertically down, Y going right horizontally) - the angle is defined as the angle wrt to the horizontal axis :

In [7]:

fig, a = mp.show_edges(np.vstack(((X/2. + .5)*mp.pe.N_X, (Y/2. + .5)*mp.pe.N_X, Theta, .1*np.ones(X.shape), np.ones(X.shape), np.zeros(X.shape) )))
t = [a.text(Y_, X_, r'(%0.1f,%0.1f), $\theta=$%0.1f' % (X_, Y_, angle*180/np.pi))  for (X_, Y_, angle)  in zip((X/2. + .5)*mp.pe.N_X, (Y/2. + .5)*mp.pe.N_X, Theta)]
a.set_xlabel('Y (columns in the image matrix - from west to east)')
_ = a.set_ylabel('X (rows in the image matrix - from north to south)')

Overlaying an image :

In [8]:

mp = SparseEdges('https://raw.githubusercontent.com/bicv/SparseEdges/master/default_param.py')
mp.pe.N = 128 # number of edges
mp.pe.figsize_edges = 9

#! defining a reference test image (see test_Image)
image = np.zeros((mp.pe.N_X, mp.pe.N_Y))
image[mp.pe.N_X//2:mp.pe.N_X//2+mp.pe.N_X//4, mp.pe.N_X//2:mp.pe.N_X//2+mp.pe.N_X//4] = 1
image[mp.pe.N_X//2:mp.pe.N_X//2+mp.pe.N_X//4, mp.pe.N_X//4:mp.pe.N_X//2] = -1

edges = np.vstack(((X/2. + .5)*mp.pe.N_X, (Y/2. + .5)*mp.pe.N_X, Theta, .1*np.ones(X.shape), np.ones(X.shape), np.zeros(X.shape) ))

fig, a = mp.show_edges(edges, image=image)
#fig, a = mp.show_edges(edges, image=im.whitening(image))
t = [a.text(Y_, X_, r'(%0.1f,%0.1f), $\theta=$%0.1f' % (X_, Y_, angle*180/np.pi))  for (X_, Y_, angle)  in zip((X/2. + .5)*mp.pe.N_X, (Y/2. + .5)*mp.pe.N_X, Theta)]
a.set_xlabel('Y (columns in the image matrix - from west to east)')
_ = a.set_ylabel('X (lines in the image matrix - from north to south)')

In [9]:

mp = SparseEdges('https://raw.githubusercontent.com/bicv/SparseEdges/master/default_param.py')
mp.set_size((32, 64))
mp.pe.N = 128 # number of edges
mp.pe.figsize_edges = 4.5

#! defining a reference test image (see test_Image)
image = np.zeros((mp.pe.N_X, mp.pe.N_Y))
image[mp.pe.N_X//2:mp.pe.N_X//2+mp.pe.N_X//4, mp.pe.N_X//2:mp.pe.N_X//2+mp.pe.N_X//4] = 1
image[mp.pe.N_X//2:mp.pe.N_X//2+mp.pe.N_X//4, mp.pe.N_X//4:mp.pe.N_X//2] = -1

edges = np.vstack(((X/2. + .5)*mp.pe.N_X-1, (Y/2. + .5)*mp.pe.N_X-1, Theta, .5*np.ones(X.shape), np.ones(X.shape), np.zeros(X.shape) ))

fig, a = mp.show_edges(edges, image=mp.whitening(image))

In [10]:

mp.set_size((256, 256))
#image = mp.imread('../../SLIP/database/lena256.png')
image = mp.imread('https://raw.githubusercontent.com/bicv/SLIP/master/database/lena256.png')

mp.pe.N = 128 # number of edges
mp.pe.figsize_edges = 9

edges = np.vstack(((X/2. + .5)*mp.pe.N_X-1, (Y/2. + .5)*mp.pe.N_X-1, Theta, .1*np.ones(X.shape), np.ones(X.shape), np.zeros(X.shape) ))

fig, a = mp.show_edges(edges, image=mp.whitening(image))

Matching Pursuit¶

checking basic edge detection¶

In [11]:

mp.set_size((128, 128))
#! defining a reference test image (see test_Image)
image = np.zeros((mp.pe.N_X, mp.pe.N_Y))
image[mp.pe.N_X//2:mp.pe.N_X//2+mp.pe.N_X//4, mp.pe.N_X//2:mp.pe.N_X//2+mp.pe.N_X//4] = 1
image[mp.pe.N_X//2:mp.pe.N_X//2+mp.pe.N_X//4, mp.pe.N_X//4:mp.pe.N_X//2] = -1
mp.imshow(image);

In [12]:

%%writefile experiment_test_MP.py
import numpy as np
from SparseEdges import SparseEdges

mp = SparseEdges('https://raw.githubusercontent.com/bicv/SparseEdges/master/default_param.py')
mp.pe.N = 128 # number of edges
mp.pe.figsize_edges = 9

#! defining a reference test image (see test_Image)
image = np.zeros((mp.pe.N_X, mp.pe.N_Y))
image[mp.pe.N_X//2:mp.pe.N_X//2+mp.pe.N_X//4, mp.pe.N_X//2:mp.pe.N_X//2+mp.pe.N_X//4] = 1
image[mp.pe.N_X//2:mp.pe.N_X//2+mp.pe.N_X//4, mp.pe.N_X//4:mp.pe.N_X//2] = -1


white = mp.pipeline(image, center=True, do_whitening=True)

import os
matname = os.path.join(mp.pe.matpath, 'experiment_test_MP.npy')
try:
    edges = np.load(matname)
except Exception:
    edges, C_res = mp.run_mp(white, verbose=True)
    try:
        os.mkdir(mp.pe.matpath)
    except Exception:
        pass
    np.save(matname, edges)   

Overwriting experiment_test_MP.py

In [13]:

%run experiment_test_MP.py

<Figure size 432x288 with 0 Axes>

In [14]:

#fig, a = mp.show_edges(edges, image=mp.whitening(image))
fig, a = mp.show_edges(edges, image=mp.dewhitening(mp.reconstruct(edges)))

In [15]:

%%writefile experiment_test_whitening.py
import numpy as np
from SparseEdges import SparseEdges
mp = SparseEdges('https://raw.githubusercontent.com/bicv/SparseEdges/master/default_param.py')
mp.pe.N = 2**9
mp.pe.MP_alpha = 1
mp.init()


image = mp.imread('https://raw.githubusercontent.com/bicv/SLIP/master/database/serre07_targets/B_N107001.jpg')

white = mp.pipeline(image, do_whitening=True)
    
import os
matname = os.path.join(mp.pe.matpath, 'experiment_test_whitening.npy')
try:
    edges = np.load(matname)
except Exception:
    edges, C_res = mp.run_mp(white, verbose=True)
    np.save(matname, edges)    


matname_MSE = os.path.join(mp.pe.matpath, 'experiment_test_whitening_MSE.npy')
try:
    MSE = np.load(matname_MSE)
except Exception:
    MSE = np.ones(mp.pe.N)
    image_rec = np.zeros_like(image)
    for i_N in range(mp.pe.N):
        MSE[i_N] =  ((white-image_rec)**2).sum()
        image_rec += mp.reconstruct(edges[:, i_N][:, np.newaxis])

    np.save(matname_MSE, MSE)     

Overwriting experiment_test_whitening.py

In [16]:

%run experiment_test_whitening.py

<Figure size 432x288 with 0 Axes>

In [17]:

#edges[0, :] = pe.N_X - edges[0,:] # HACK
#edges[2, :] = - edges[2,:] # HACK
#edges[:, 119] = edges[:, 118] # HACK
fig_width_pt = 318.670  # Get this from LaTeX using \showthe\columnwidth
inches_per_pt = 1.0/72.27               # Convert pt to inches
fig_width = fig_width_pt*inches_per_pt  # width in inches
mp.pe.figsize_edges = .382 * fig_width
fig, a = mp.show_edges(edges, image=image, color='brown', show_phase=False) #, pedestal=.0, gamma=.3
#import os
#fig.savefig(os.path.join(mp.pe.figpath, 'SparseLets_A.pdf'))

In [18]:

#! recontructing the image from the edges
image_rec = mp.reconstruct(edges)
print('MSE-W = ', ((white-image_rec)**2).sum()/((white)**2).sum())

MSE-W =  0.3351719916880489

In [19]:

fig, a = mp.show_edges(edges, image=image_rec*1., color='brown')

Recontructing the original image by dewhitening the reconstructed image:

In [20]:

fig, a = mp.show_edges(edges, image=mp.dewhitening(image_rec)*1., color='brown')

In [21]:

MSE = np.load(os.path.join(mp.pe.matpath, 'experiment_test_whitening_MSE.npy'))
print(MSE.shape, mp.pe.N)

(512,) 512

In [22]:

# checking the quick computation of the MSE in MP
MSE_0 = (white**2).sum()
print('stats on white:', white.mean(), white.std(), np.sqrt(MSE_0), np.sqrt(MSE[0]))
print ('mp.pe.MP_alpha=', mp.pe.MP_alpha)    

MSE_MP = np.ones(mp.pe.N)
MSE_MP[1:] = 1. - np.cumsum(edges[4, :-1]**2) * (2 -  mp.pe.MP_alpha)/mp.pe.MP_alpha / MSE_0

plt.figure(figsize=(12,6))
plt.subplot(111)
plt.plot(np.sqrt(MSE/MSE[0]), 'ro', label='true', alpha=.2)
plt.plot(np.sqrt(MSE_MP), 'g--', label='MP-vec', lw=2)
plt.xlim([0, mp.pe.N])
plt.ylim([0, 1])
plt.xlabel('# atoms')
plt.ylabel('MSE')
_ = plt.legend()

stats on white: -0.00020648729955733365 0.0902248428093429 23.0976202472936 23.0976202472936
mp.pe.MP_alpha= 1

In [23]:

%%writefile experiment_test_nowhitening.py
import numpy as np
from SparseEdges import SparseEdges
mp = SparseEdges('https://raw.githubusercontent.com/bicv/SparseEdges/master/default_param.py')
mp.pe.N = 256
mp.pe.do_whitening = False

#image = mp.imread('https://raw.githubusercontent.com/bicv/SLIP/master/database/serre07_targets/B_N107001.jpg')
image = mp.imread('../../SLIP/database/serre07_targets/B_N107001.jpg')


mp.pe.figsize_edges = 9
image = mp.pipeline(image, do_whitening=False)


import os
matname = os.path.join(mp.pe.matpath, 'experiment_test_nowhitening.npy')
try:
    edges = np.load(matname)
except Exception:
    edges, C_res = mp.run_mp(image, verbose=True)
    np.save(matname, edges)    
    

Overwriting experiment_test_nowhitening.py

In [24]:

%run experiment_test_nowhitening.py

<Figure size 432x288 with 0 Axes>

In [25]:

#! recontructing the image from the edges
image_ = mp.reconstruct(edges)
print('MSE = ', ((image-image_)**2).sum()/((image)**2).sum())

MSE =  0.08428340351792607

In [26]:

#! recontructing the image from the edges
image_ = mp.reconstruct(edges)
print('WMSE = ', ((mp.whitening(image-image_))**2).sum()/((mp.whitening(image))**2).sum())

WMSE =  0.6103895965836417

In [27]:

fig, a = mp.show_edges(edges, image=image_*1., show_mask=True, color='brown')

some book keeping for the notebook¶

In [28]:

%load_ext watermark
%watermark

2018-07-13T17:10:45+02:00

CPython 3.7.0
IPython 6.4.0

compiler   : Clang 9.1.0 (clang-902.0.39.2)
system     : Darwin
release    : 17.6.0
machine    : x86_64
processor  : i386
CPU cores  : 4
interpreter: 64bit

In [29]:

%load_ext version_information
%version_information numpy, scipy, matplotlib, sympy, pillow, imageio

Out[29]:

Software	Version
Python	3.7.0 64bit [Clang 9.1.0 (clang-902.0.39.2)]
IPython	6.4.0
OS	Darwin 17.6.0 x86_64 i386 64bit
numpy	1.14.5
scipy	1.1.0
matplotlib	2.2.2
sympy	1.2
pillow	5.2.0
imageio	2.3.0
Fri Jul 13 17:10:46 2018 CEST

In [30]:

%cd -q ../notebooks