In [5]:

    

import amitgroup as ag


    

In [6]:

    

import numpy as np
a = range(10)
img = ag.io.load_mnist("training",[4], selection= slice(100),return_labels=False)


    

In [7]:

    

import amitgroup.plot as gr


    

In [8]:

    

gr.images(img)


    

In [9]:

    

img2 = ag.io.load_mnist("training",[9], selection= slice(100),return_labels=False)
gr.images(img2)


    

In [10]:

    

img3 = ag.io.load_mnist("training",[7], selection= slice(100),return_labels=False)
gr.images(img3)


    

In [11]:

    

img4 = ag.io.load_mnist("training",[1], selection= slice(100),return_labels=False)
gr.images(img4)


    

In [12]:

    

import amitgroup as ag
import pnet


    

In [ ]:

    

edgeLayer = pnet.EdgeLayer(k=5, radius=1, spread='orthogonal', minimum_contrast=0.05)
#pnet.IntensityThresholdLayer(),


    

In [98]:

    

partsLayer = pnet.PartsLayer(20, (6, 6), settings=dict(outer_frame=0,
                                          threshold=150,
                                          samples_per_image=200,
                                          max_samples=100000,
                                          min_prob=0.005,
                                          ))


    

In [99]:

    

a = range(10)
img = ag.io.load_mnist("training",a, selection= slice(1000),return_labels=False)


    

In [94]:

    

edgeLayer.train(img)


    

In [95]:

    

imgEdges = edgeLayer.extract(img)


    

In [96]:

    

imgEdges.shape


    

    
Out[96]:





(1000, 28, 28, 8)

In [100]:

    

patches, originalPatches = partsLayer._get_patches(imgEdges,img)


    

In [23]:

    

type(patches)


    

    
Out[23]:





tuple

In [101]:

    

originalPatches.shape


    

    
Out[101]:





(0,)

In [89]:

    

gr.images(originalPatches[0:100])


    

    






<matplotlib.figure.Figure at 0x11826b910>

In [90]:

    

partsLayer.train_from_samples(patches,originalPatches)


    

    




---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-90-60133cdf0aa9> in <module>()
----> 1 partsLayer.train_from_samples(patches,originalPatches)

/Users/jiajunshen/Documents/Research/partsNet/pnet/parts_layer.pyc in train_from_samples(self, patches, patches_original)
     63         from pnet.bernoullimm import BernoulliMM
     64         min_prob = self._settings.get('min_prob', 0.01)
---> 65         flatpatches = patches.reshape((patches.shape[0], -1))
     66 
     67         if 1:

ValueError: total size of new array must be unchanged

In [ ]:

    

extractedFeature,num = partsLayer.extract(patches)


    

In [ ]:

    

meanPart = np.array([originalPatches[extractedFeature[:,0,0,0]==k].mean(axis = 0) for k in range(partsLayer._num_parts)])


    

In [81]:

    

totalFeature.shape


    

    
Out[81]:





(10838, 1, 1, 1)

In [75]:

    

originalPatches[extractedFeature[:,0,0,0]==2].shape


    

    
Out[75]:





(0, 6, 6)

In [ ]:

    

meanPart.shape


    

In [ ]:

    

gr.images(meanPart)


    

In [ ]: