An example MP run.
To run:
$ python experiment_example.py
To remove cache:
$ rm -fr **/example*
"""
__author__ = "(c) Laurent Perrinet INT - CNRS"
import numpy as np
from SparseEdges import SparseEdges
mp = SparseEdges("https://raw.githubusercontent.com/bicv/SparseEdges/master/default_param.py")
mp.N = 128
image = mp.imread("https://raw.githubusercontent.com/bicv/SparseEdges/master/database/lena256.png")
name = "example"
image = mp.normalize(image, center=True)
# print image.mean(), image.std()
import os
matname = os.path.join(mp.pe.matpath, name + ".npy")
try:
edges = np.load(matname)
except:
edges, C_res = mp.run_mp(image, verbose=True)
from __future__ import division, print_function
"""
$ ipython experiment_fig-efficiency.py
rm -fr **/efficiency_* **/**/efficiency_*
"""
import numpy as np
from SparseEdges import SparseEdges
FORMATS = ['pdf', 'eps']
threshold = None # classical plots
threshold = .3 # plot L0 sparseness obtained when reaching this threshold
mp = SparseEdges('https://raw.githubusercontent.com/bicv/SparseEdges/master/default_param.py')
def init_mp():
mp = SparseEdges('https://raw.githubusercontent.com/bicv/SparseEdges/master/default_param.py')
mp.pe.seed = 42
mp.pe.N_image = 60
mp.pe.datapath = '../../SLIP/database/'
return mp
FORMATS = ['pdf', 'eps']
#FORMATS = ['png']
import matplotlib
matplotlib.use('Agg')
#matplotlib.rcParams.update({'font.size': 18, 'font.family': 'STIXGeneral', 'mathtext.fontset': 'stix'})
matplotlib.rcParams.update({'text.usetex': False})
"""
__author__ = "(c) Laurent Perrinet INT - CNRS"
import numpy as np
import matplotlib
matplotlib.use("Agg") # agg-backend, so we can create figures without x-server (no PDF, just PNG etc.)
from SparseEdges import SparseEdges
FORMATS = ['pdf', 'eps']
# TODO: here, we are more interested in the processing of the database, not the comparison - use the correct function
# TODO : annotate the efficiency of different LogGabor bases (RMSE?)
# TODO: make a circular mask to avoid border effects coming with whitening...
#! comparing databases
#!--------------------
mp = SparseEdges('https://raw.githubusercontent.com/bicv/SparseEdges/master/default_param.py')
mp.N = 128
mp.pe.datapath = '../../SLIP/database/'
mp.process('testing_vanilla')
# TODO: CRF
mp.process('testing_noise', noise=mp.pe.noise)
mp.process('testing_vanilla', name_database='serre07_targets')
# TODO : make an experiment showing that using scale does not bring much
mps, experiments = [], []
v_alpha = np.linspace(0.3, 1., 9)
for MP_alpha in v_alpha:
mp = SparseEdges('https://raw.githubusercontent.com/bicv/SparseEdges/master/default_param.py')
mp.N = 128
mp.pe.datapath = '../../SLIP/database/'
mp.pe.MP_alpha = MP_alpha
$ python experiment_fig-sparselets.py ./figures
$ rm -fr **/SparseLets* **/**/SparseLets*
"""
__author__ = "(c) Laurent Perrinet INT - CNRS"
from SparseEdges import SparseEdges
FORMATS = ['pdf', 'eps']
mps = []
sizes = [16, 32, 64, 128, 256]
N_image = 32
N = 1024
for size, size_str in zip(sizes, ['_016', '_032', '_064', '_128', '']):
mp = SparseEdges('https://raw.githubusercontent.com/bicv/SparseEdges/master/default_param.py')
mp.pe.seed = 42
mp.pe.datapath = '../../SLIP/database/'
mp.set_size((size, size))
downscale_factor = sizes[-1]/size # > 1
mp.pe.N_image = int(N_image*downscale_factor)
mp.pe.N = int(N/downscale_factor**2)
mp.init()
mp.process('SparseLets' + size_str)
mps.append(mp)
import matplotlib.pyplot as plt
fig_width_pt = 600 # 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
fig = plt.figure(figsize=(fig_width, fig_width/1.618))
from __future__ import division, print_function
"""
$ python experiment_fig-firstorder.py
rm -fr **/prior_* **/**/prior_*
rm -fr **/prior_vanilla* **/**/prior_vanilla*
rm -fr **/prior_vanilla_noise_* **/**/prior_vanilla_noise_*
rm -fr **/prior_firstorder* **/**/prior_firstorder*
"""
import numpy as np
from SparseEdges import SparseEdges
for name_database in ['serre07_distractors']:#, 'serre07_distractors_urban', 'laboratory']:
mp = SparseEdges('https://raw.githubusercontent.com/bicv/SparseEdges/master/default_param.py')
mp.pe.datapath = '../../SLIP/database/'
mp.pe.seed = 21341353 # this ensures that all image lists are the same for the different experiments
mp.pe.N_image = 20
mp.pe.N = 1024
# control experiment
#mp.theta = np.linspace(-np.pi/2, np.pi/2, mp.n_theta+1)[1:]
imageslist, edgeslist, RMSE = mp.process(exp='prior_vanilla', name_database=name_database)
imageslist_noise, edgeslist_noise, RMSE_noise = mp.process(exp='prior_vanilla_noise_' + str(mp.pe.noise).replace('.', '_'), name_database=name_database, noise=mp.pe.noise)
if True:#try:
six, N, N_image = edgeslist.shape
# first-order prior
v_hist, v_theta_edges = mp.histedges_theta(edgeslist, display=False)
import numpy as np
from SparseEdges import SparseEdges
mp = SparseEdges('https://raw.githubusercontent.com/bicv/SparseEdges/master/default_param.py')
mp.N = 128
image = mp.imread('https://raw.githubusercontent.com/bicv/SLIP/master/database/serre07_targets/B_N107001.jpg')
mp.pe.figsize_edges = 9
image = mp.normalize(image, center=True)
#! trying now using no whitening of the image
mp.pe.do_whitening = False
import os
matname = os.path.join(mp.pe.matpath, 'experiment_test_nowhitening.npy')
matname_RMSE = os.path.join(mp.pe.matpath, 'experiment_test_nowhitening_RMSE.npy')
try:
edges = np.load(matname)
except:
edges, C_res = mp.run_mp(image, verbose=True)
np.save(matname, edges)
fig, a = mp.show_edges(edges, image=image, mask=True)
import numpy as np
from SparseEdges import SparseEdges
mp = SparseEdges('https://raw.githubusercontent.com/bicv/SparseEdges/master/default_param.py')
mp.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
import os
matname = os.path.join(mp.pe.matpath, 'experiment_test_MP.npy')
try:
edges = np.load(matname)
except:
edges, C_res = mp.run_mp(image, verbose=False)
np.save(matname, edges)
fig, a = mp.show_edges(edges, image=mp.whitening(image))
# -*- coding: utf8 -*-
from __future__ import division, print_function
"""
$ python experiment_retina_sparseness.py
rm -fr **/retina_sparseness* **/**/retina_sparseness*
"""
import numpy as np
from SparseEdges import SparseEdges
mps = []
for name_database in ["serre07_distractors"]: # , 'serre07_distractors_urban', 'laboratory']:
mp = SparseEdges("https://raw.githubusercontent.com/bicv/SparseEdges/master/default_param.py")
mp.pe.datapath = "../../SLIP/database/"
mp.pe.N_image = 100
mp.pe.do_mask = True
mp.pe.N = 2 ** 12
mp.pe.n_theta = 1
mp.pe.B_theta = np.inf
mp.pe.line_width = 0
mp.init()
# normal experiment
imageslist, edgeslist, RMSE = mp.process(exp="retina_sparseness", name_database=name_database)
mps.append(mp)
# control experiment
if True:
mp.pe.MP_alpha = np.inf
mp.init()
