def choose_samples(indices=[0, 3, 6, 16]):
import plankton_vis1
samples = plankton_vis1.loadSamplePlanktons(20) # 100,1,28,28
if False:
for i, sample in enumerate(samples):
print 'Showing sample ', i
plt.imshow(sample[0], cmap=cm.Greys_r)
plt.show()
return samples[indices]
def choose_samples(indices=[0,3,6,16]):
import plankton_vis1
samples = plankton_vis1.loadSamplePlanktons(20) # 100,1,28,28
if False:
for i, sample in enumerate(samples):
print 'Showing sample ',i
plt.imshow(sample[0], cmap = cm.Greys_r)
plt.show()
return samples[indices]
def Find_plankton(model_name="plankton_conv_visualize_model.pkl.params"):
"""
Find plankton that activates the given layers most
"""
which_layer = 2
import plankton_vis1
samples = plankton_vis1.loadSamplePlanktons(numSamples=3000)
print 'Dimension of Samples', np.shape(samples)
Net = DeConvNet(model_name)
#kernel_list = [ 2,23,60,12,45,9 ]
kernel_list = range(48,64)
num_of_maximum = 9
Heaps = findmaxactivation( Net, samples, num_of_maximum, kernel_list, which_layer=which_layer)
bigbigmap = None # what is this?
for kernel_index in Heaps:
print 'dealing with',kernel_index,'th kernel'
heap = Heaps[kernel_index]
this_sams = []
this_Deconv = []
for pairs in heap:
this_sam = pairs.sam
this_sams.append( this_sam.reshape([NUM_C,MAX_PIXEL,MAX_PIXEL]) )
this_Deconv.append( Net.DeConv( this_sam, kernel_index, which_layer=which_layer ).reshape([NUM_C,MAX_PIXEL,MAX_PIXEL]) )
this_sams = np.array( this_sams )
this_sams = np.transpose( this_sams, [ 1, 0, 2, 3 ])
this_sams = this_sams.reshape( [ NUM_C, 9, MAX_PIXEL*MAX_PIXEL ])
this_sams = tuple( [ this_sams[0] for i in xrange(3)] + [None] )
this_Deconv = np.array( this_Deconv )
this_Deconv = np.transpose( this_Deconv, [ 1, 0, 2, 3 ])
this_Deconv = this_Deconv.reshape( [ NUM_C, num_of_maximum, MAX_PIXEL*MAX_PIXEL ])
this_Deconv = tuple( [ this_Deconv[0] for i in xrange(3)] + [None] )
this_map = tile_raster_images( this_sams, img_shape = (MAX_PIXEL,MAX_PIXEL), tile_shape = (3,3),
tile_spacing=(1, 1), scale_rows_to_unit_interval=True,
output_pixel_vals=True)
this_Deconv = tile_raster_images( this_Deconv, img_shape = (MAX_PIXEL,MAX_PIXEL), tile_shape = (3,3),
tile_spacing=(1, 1), scale_rows_to_unit_interval=True,
output_pixel_vals=True)
this_pairmap = np.append( this_map, this_Deconv, axis = 0)
if bigbigmap == None:
bigbigmap = this_pairmap
segment_line = 255*np.ones([bigbigmap.shape[0],1,4],dtype='uint8')
else:
bigbigmap = np.append(bigbigmap, segment_line, axis = 1)
bigbigmap = np.append(bigbigmap, this_pairmap, axis = 1)
plt.imshow(bigbigmap)
plt.show()
def main(model_name="plankton_model0.pkl.params",
rotate=False, bScalar=True,
which_layer=2, numSamples='All'):
import plankton_vis1
samples = plankton_vis1.loadSamplePlanktons(numSamples, rotate)
print 'Dimension of Samples', np.shape(samples)
activations = findActivations(model_name, samples, which_layer,
allFeats=False, bScalar=True)
y_labels = plankton_utils.getY_label_int(which_data='train')
activation_dict = separateActivationsByClass(activations, y_labels)
calculateAverageActivations(activation_dict, which_layer)
plotFeatures(activation_dict, which_layer)
def main(model_name="plankton_model0.pkl.params",
rotate=False,
bScalar=True,
which_layer=2,
numSamples='All'):
import plankton_vis1
samples = plankton_vis1.loadSamplePlanktons(numSamples, rotate)
print 'Dimension of Samples', np.shape(samples)
activations = findActivations(model_name,
samples,
which_layer,
allFeats=False,
bScalar=True)
y_labels = plankton_utils.getY_label_int(which_data='train')
activation_dict = separateActivationsByClass(activations, y_labels)
calculateAverageActivations(activation_dict, which_layer)
plotFeatures(activation_dict, which_layer)
def displayClassSamples(saveIndividualSamples=False):
# from bc01 to cb01
samples = loadSamplePlanktons(30000, rotate=False)
print 'samples shape', samples.shape #e.g. (30000,1,28,28)
y_labels = plankton_utils.getY_label_int(which_data='train')[:30000]
# construct list of indices for each label
class_indices = [None]*121
for _index, y_label in enumerate(y_labels):
#print "label %d _index = %d" % (y_label, _index)
if class_indices[y_label] == None:
class_indices[y_label] = [_index]
else:
class_indices[y_label].append(_index);
#print class_indices[y_label]
for i, this_class_indices in enumerate(class_indices):
if len(this_class_indices) < 4:
print 'Warning! Increase sample size! (Found less than 4 samples for class ', i
return
if saveIndividualSamples:
''' Saving sample image for each class '''
#print 'this_class_indices', this_class_indices
shape = samples[this_class_indices[0]].shape
#print 'shape of sample', shape #e.g. (1,28,28)
for class_num, this_class_indices in enumerate(class_indices):
tile4 = None
for i in range(2):
tile2 = None
for j in range(2):
if tile2 == None:
tile2 = np.reshape(samples[this_class_indices[2*i+j]], (shape[1],shape[2]))
else:
tile2 = np.concatenate( (tile2,
np.reshape(samples[this_class_indices[2*i+j]], (shape[1],shape[2]))
),
axis=0)
if tile4 == None:
tile4 = tile2
else:
tile4 = np.concatenate( (tile4, tile2), axis=1)
print 'class_num', class_num ,'shape of tile4', tile4.shape
fname = 'results/classSamples_grey/' + str(class_num) + ".png"
if False:
from PIL import Image
newIm = 255.*(tile4)/2. + 127.
im = Image.fromarray(newIm)
im = im.convert('RGB')
print 'Saving class num=', class_num
im.save(fname)
else:
plt.imshow(newIm, cmap = cm.Greys_r)
plt.axis('off')
plt.show()
plt.savefig(fname)
bigbigmap = None
for row in range(11):
bigmap = None
for col in range(11):
class_index = 11*row+col
print 'Class Index', class_index
print "Indices with this class", class_indices[class_index][:4]
class_samples = samples[class_indices[class_index][:4]]
print "class_samples shape", class_samples.shape
'''
for pairs in heap:
this_sam = pairs.sam
this_sams.append( this_sam.reshape([NUM_C,MAX_PIXEL,MAX_PIXEL]) )
this_Deconv.append( Net.DeConv( this_sam, kernel_index, which_layer=which_layer ).reshape([NUM_C,MAX_PIXEL,MAX_PIXEL]) )
'''
this_sams = np.transpose( class_samples, [ 1, 0, 2, 3 ])
this_sams = this_sams.reshape( [ NUM_C, 4, MAX_PIXEL*MAX_PIXEL ])
this_sams = tuple( [ this_sams[0] for i in xrange(3)] + [None] )
this_map = tile_raster_images( class_samples, img_shape = (MAX_PIXEL,MAX_PIXEL), tile_shape = (2,2),
tile_spacing=(1, 1), scale_rows_to_unit_interval=True,
output_pixel_vals=True)
if bigmap == None:
bigmap = this_map
segment_line = 255*np.ones([bigmap.shape[0],1],dtype='uint8')
print 'this_map dimension', this_map.shape
print 'segment_line dimension', segment_line.shape
else:
bigmap = np.append(bigmap, segment_line, axis = 1)
bigmap = np.append(bigmap, this_map, axis = 1)
if bigbigmap == None:
bigbigmap = bigmap
else:
bigbigmap = np.append(bigbigmap, 255*np.ones((1, bigmap.shape[1]), dtype='uint8'), axis=0)
bigbigmap = np.append(bigbigmap, bigmap, axis=0)
print 'shape of big big map', bigbigmap.shape
plt.imshow(bigbigmap,cmap = cm.Greys_r)
plt.xticks(np.arange(11))
plt.yticks(np.arange(11))
plt.show()