In [1]:

    

DIR = "F:/data/101_ObjectCategories/"


    

In [2]:

    

import sys;sys.path.append("C:/Users/jmenard/learning_smoke/scripts")


    

In [126]:

    

import pandas as pd
import numpy as np
import os
import matplotlib.pyplot as plt
from skimage.transform import resize as imresize 
import skimage.io as skio
import skimage.transform 
import pywt
from sklearn.linear_model import Lasso


    

In [127]:

    

os.listdir(DIR)[0:10]


    

    
Out[127]:





['accordion',
 'airplanes',
 'anchor',
 'ant',
 'BACKGROUND_Google',
 'barrel',
 'bass',
 'beaver',
 'binocular',
 'bonsai']

In [128]:

    

DIR_LOC = DIR+"bonsai/"


    

In [129]:

    

os.listdir(DIR_LOC)[0:10]


    

    
Out[129]:





['image_0001.jpg',
 'image_0002.jpg',
 'image_0003.jpg',
 'image_0004.jpg',
 'image_0005.jpg',
 'image_0006.jpg',
 'image_0007.jpg',
 'image_0008.jpg',
 'image_0009.jpg',
 'image_0010.jpg']

In [7]:

    

len( os.listdir(DIR_LOC) )


    

    
Out[7]:





128

In [8]:

    

_img = skio.imread(DIR_LOC+"image_0001.jpg")


    

In [9]:

    

print(
    type(_img),
    _img.shape,
    _img.dtype

)


    

    




<class 'numpy.ndarray'> (300, 280, 3) uint8

In [84]:

    

plt.imshow(_img[:,:,0])


    

    
Out[84]:





<matplotlib.image.AxesImage at 0x1ea183b6400>






    

In [66]:

    

_img.shape


    

    
Out[66]:





(300, 280, 3)

In [113]:

    

_img_flat = _img[:,:,0].flatten()


    

In [114]:

    

_img_flat.shape


    

    
Out[114]:





(84000,)

In [115]:

    

# Get fft coefficients:
_s = np.fft.fft(_img_flat)
re_s = np.real(_s)
im_s = np.imag(_s)


    

In [116]:

    

plt.plot(re_s, im_s,linestyle="None",marker=".")


    

    
Out[116]:





[<matplotlib.lines.Line2D at 0x1ea1893cf28>]






    

In [117]:

    

_s_sorted = np.sort(abs(_s))


    

In [118]:

    

thresh_5per = _s_sorted[int(0.05*300*280)]


    

In [119]:

    

big_idxs = _s > thresh_5per
_s_cut   = np.multiply(_s, big_idxs)


    

In [120]:

    

# Get fft coefficients:
_s_cut    = np.fft.fft(_img_flat)
re_s_sort = np.real(_s_cut)
im_s_sort = np.imag(_s_cut)
plt.plot(re_s_sort, im_s_sort,linestyle="None",marker=".")


    

    
Out[120]:





[<matplotlib.lines.Line2D at 0x1ea189c71d0>]






    

In [121]:

    

inv_s = np.fft.ifft(_s_cut)


    

In [122]:

    

inv_s.shape


    

    
Out[122]:





(84000,)

In [123]:

    

rec_ = inv_s.reshape(300,280)


    

In [124]:

    

plt.imshow( np.real(rec_) )
plt.title("Taking top 5% of fourier coefs")


    

    
Out[124]:





Text(0.5,1,'Taking top 5% of fourier coefs')






    

In [125]:

    

plt.imshow(_img[:,:,0])
plt.title("Original")


    

    
Out[125]:





Text(0.5,1,'Original')






    

In [ ]: