def get_overfeat_features(imgs, weights_path, typ, layer=None, cache=None):
"""Returns features at layer for given image(s) from OverFeat model.
Small (fast) network: 22 layers
Large (accurate) network: 25 layers
Args:
imgs: Iterable of images each of shape (h,w,c)
weights_path: Path to the OverFeat weights
typ: 0 for small, 1 for large version of OverFeat
layer: The layer to extract features from
cache: Dict containing descs/other cached values
"""
if cache is None:
cache = {}
if 'overfeat_descs' not in cache:
# Initialize network
print('Loading OverFeat ({}) model...'.format(typ))
overfeat.init(weights_path, typ)
# Determine feature layer if none specified
if layer is None:
if overfeat.get_n_layers() == 22: # small
layer = 19 # 16 also recommended
else: # large
# Layer used by Zhang et al.
layer = 22
# Determine resize dim
if typ == 0:
resize = 231 # small network
else:
resize = 221 # large network
# Allocate for feature descriptors
descs = []
# Run images through network
print('Running images through OverFeat, extracting features '
'at layer {}.'.format(layer))
for idx, img in enumerate(imgs):
if (idx + 1) % 100 == 0:
print('Processing image {}...'.format(idx + 1))
# Preprocess image
img = overfeat_preprocess(img, resize)
# Run through model
_ = overfeat.fprop(img)
# Retrieve feature output
desc = overfeat.get_output(layer)
descs.append(desc)
# Free network
overfeat.free()
# NumPy-ify
descs = np.asarray(descs)
cache.update(overfeat_descs=descs)
else:
descs = cache['overfeat_descs']
return descs, cache
def extract_feat(self, imgs, ftype):
'''
extract features
'''
if ftype is 'hog':
from skimage.feature import hog
L = np.sqrt(len(imgs[0])).astype(int)
X_im = [imresize(arr.reshape((L, L)), (64, 64)) for arr in imgs]
pool = Pool(processes=8)
X = pool.map(hog, X_im)
pool.close()
return np.asarray(X)
elif ftype is 'overfeat':
overfeat.init('OverFeat/data/default/net_weight_0', 1)
L = np.sqrt(len(imgs[0])).astype(int)
imgs_color = [imresize(arr.reshape((L, L)), (231, 231)) for arr in imgs]
if len(imgs_color[0].shape) != 3:
cmap = pl.get_cmap('jet')
imgs_color = [np.delete(cmap(im/255.), 3, 2) for im in imgs_color]
imgs_roll = [im.transpose((2, 0, 1)).astype(np.float32) for im in imgs_color]
feats = np.zeros((len(imgs_roll), 4096), dtype = float)
for i in range(len(imgs_roll)):
b = overfeat.fprop(imgs_roll[i])
f22 = overfeat.get_output(22)
f22 = np.asarray(f22).squeeze().astype(np.float)
feats[i, :] = f22
return feats
elif ftype is 'pix':
return imgs
else:
raise NameError('{0} is not implemented!'.format(ftype))
from sklearn.externals import joblib
from scipy.ndimage import imread
from scipy.misc import imresize
LAYER = 6
X = None
L = []
with open('data/train_outputs.csv', 'rb') as f:
reader = csv.reader(f, delimiter=',')
next(reader, None)
for data in reader:
L.append(data)
# quick network: layers 17, 19
# slow network: layers 20, 22
overfeat.init('/Users/npow/code/OverFeat/data/default/net_weight_0', 0)
def extract_features(layer, file_name):
global X
image = imread(file_name)
image.resize((image.shape[0], image.shape[1], 1))
# overfeat expects rgb, so replicate the grayscale values twice
image = np.repeat(image, 3, 2)
image = imresize(image, (231, 231)).astype(np.float32)
# numpy loads image with colors as last dimension, so transpose tensor
h = image.shape[0]
w = image.shape[1]
if filename == ".directory":
continue
image = imread(os.path.join(directory, filename)).astype(numpy.float32)
# numpy loads image with colors as last dimension, transpose tensor
h = image.shape[0]
w = image.shape[1]
c = image.shape[2]
image = image.reshape(w * h, c)
image = image.transpose()
image = image.reshape(c, h, w)
# run overfeat on the image
overfeat.fprop(image)
features = overfeat.get_output(19)
d[filename] = features.copy()
pickle.dump(d, open(dirname + "_full.txt", "w"))
return d
overfeat.init('/home/juanjo/U/overfeat/data/default/net_weight_0', 0)
rootDir = "/home/juanjo/U/Búsqueda por Contenido en Imágenes y Videos/Proyecto/ResizedImages_6/"
# dTotal = {}
for subdir in os.listdir(rootDir):
print subdir
if os.path.exists(subdir + "_full.txt"):
continue
d = getDictOfFeatures(os.path.join(rootDir, subdir), subdir)
# dTotal[subdir] = d
# pickle.dump(dTotal, open("features_full.txt", "w"))
# resize and crop into a 231x231 image
h0 = image.shape[0]
w0 = image.shape[1]
d0 = float(min(h0, w0))
image = image[int(round((h0-d0)/2.)):int(round((h0-d0)/2.)+d0),
int(round((w0-d0)/2.)):int(round((w0-d0)/2.)+d0), :]
image = imresize(image, (231, 231)).astype(numpy.float32)
# numpy loads image with colors as last dimension, transpose tensor
h = image.shape[0]
w = image.shape[1]
c = image.shape[2]
image = image.reshape(w*h, c)
image = image.transpose()
image = image.reshape(c, h, w)
print "Image size :", image.shape
# initialize overfeat. Note that this takes time, so do it only once if possible
overfeat.init('../../data/default/net_weight_0', 0)
# run overfeat on the image
b = overfeat.fprop(image)
# display top 5 classes
b = b.flatten()
top = [(b[i], i) for i in xrange(len(b))]
top.sort()
print "\nTop classes :"
for i in xrange(5):
print(overfeat.get_class_name(top[-(i+1)][1]))
# OverFeat has 22 layers, including original image
num_layers = 22
# For filename purposes
layer = 'all'
layer = 10
if layer == 'all':
# Put all layers into one stacked confusion matrix
confusion_matrix = np.zeros((num_layers, len(training_images), len(testing_images)))
else:
# Make the confusion matrix for a single layer
confusion_matrix = np.zeros((len(training_images), len(testing_images)))
overfeat.init(overfeat_root + 'data/default/net_weight_0', 0)
for i in range(len(training_images)):
print("Training Image %s of %s" % (i, len(training_images)))
if smush:
image = smush_overfeat_images(training_images[i])
else:
image = load_overfeat_image(training_images[i])
b = overfeat.fprop(image)
if layer == 'all':
# Calculate features for all layers at once
features = []
from datetime import datetime
from multiprocessing import Process, Pool
from sensor_msgs.msg import Image
from rospy.numpy_msg import numpy_msg
from scipy.misc import imresize, imsave
from sklearn import svm
from sklearn.cross_validation import StratifiedKFold, cross_val_score
from skimage.feature import hog
OVERFEAT_DIM = 231
KINECT_WIDTH = 640
KINECT_HEIGHT = 480
# initialize overfeat. 0 = fast net, 1 = accurate net.
overfeat.init("../data/weights/net_weight_0", 0)
# list of unique classes as strings.
# indices correspond to the 'target' of each 'classification'.
classes = []
# list of training samples.
training = []
# each Sample is features data and the target of a classification.
Sample = namedtuple('Sample',
['features', 'depth', 'classification', 'time', 'image'])
# will train svm classifier on features extracted using overfeat
svm_clf = svm.LinearSVC()
def fit(self, X=None, y=None):
import overfeat # soft dep
overfeat.init(self.pretrained_params, self.network_size)
return self
from PIL import Image
import numpy
import skimage
from skimage.feature import daisy, hog
import subprocess
import overfeat
import numpy
from scipy.ndimage import imread
from scipy.misc import imresize
import database_file
import time
from find_obj import init_feature
from asift import affine_detect
print "Initializing overfeat"
overfeat.init('overfeat/overfeat/data/default/net_weight_0', 0)
#overfeat_initialized = False
def extract_overfeat(image_path):
# if overfeat_initialized == None or not overfeat_initialized:
# overfeat_initialized = True
print "Overfeat: ", image_path
image = imread(image_path)
print "Image shape: ", image.shape
if len(image.shape) == 2 or image.shape[2] == 2:
image = skimage.color.gray2rgb(image)
elif image.shape[2] == 4:
image_rgb = numpy.zeros((image.shape[0], image.shape[1], 3),
numpy.uint8)
def initialize(self):
# initialize overfeat. Note that this takes time, so do it only once if possible
if self.debug == True:
pdb.set_trace()
overfeat.init('../../data/default/net_weight_'+str(self.network), self.network)
w0 = image.shape[1]
d0 = float(min(h0, w0))
h1 = int(round(231*h0/d0))
w1 = int(round(231*w0/d0))
image = imresize(image, (h1, w1)).astype(numpy.float32)
image = image[int(round((h0-d0)/2.)):int(round((h0-d0)/2.)+231),
int(round((w0-d0)/2.)):int(round((w0-d0)/2.)+231), :]
# numpy loads image with colors as last dimension, transpose tensor
h = image.shape[0]
w = image.shape[1]
c = image.shape[2]
image = image.reshape(w*h, c)
image = image.transpose()
image = image.reshape(c, h, w)
print "Image size :", image.shape
# initialize overfeat. Note that this takes time, so do it only once if possible
overfeat.init('../../data/default/net_weight_0', 0)
# run overfeat on the image
b = overfeat.fprop(image)
# display top 5 classes
b = b.flatten()
top = [(b[i], i) for i in xrange(len(b))]
top.sort()
print "\nTop classes :"
for i in xrange(5):
print(overfeat.get_class_name(top[-(i+1)][1]))
# numpy loads image with colors as last dimension, transpose tensor h = image.shape[0] w = image.shape[1] c = image.shape[2] image = image.reshape(w*h, c) image = image.transpose() image = image.reshape(c, h, w) except: return None return image def print_features(image_path, features): line = image_path + '\t' + '\t'.join(map(str,features)) print(line) # initialize overfeat. Note that this takes time, so do it only once if possible overfeat.init(network_path, 0) for image_path in sys.stdin: image_path = image_path[:-1] # remove \n char image = read_image(image_path) # run overfeat on the image features = [] if not image is None: b = overfeat.fprop(image) features = overfeat.get_output(layer_index).flatten() print_features(image_path, features)
'''train and test scikit.svm on photo features extracted using overfeat.'''
# read in and reshape photos
photo_files = get_photo_files(get_photo_path())
photos = [read_in_photo(path) for path in photo_files]
# split photos into training and test
test_photo = photos.pop(-2)
train_photos = photos
# (TODO don't hard code targets)
test_target = np.array([1])
train_targets = np.array([0,0,0,0,0,0,0,0,1,1,1,1,1,1,1])
# TODO make this cleaner
photo_files.pop(-2)
# initialize overfeat weights. fast net: 0. large net: 1.
overfeat.init("./data/default/net_weight_0", 0)
# concurrently extract photo features and predictions using overfeat
pool = Pool()
likelihoods, train_features = [np.array(tup) for tup in zip(*pool.map(process_through_net, train_photos))]
pool.close()
pool.join()
# print overfeat predictions
print_overfeat_predictions(photo_files, likelihoods)
# train svm on extracted photo features
classifier = svm.SVC()
classifier.fit(train_features, train_targets)
# test classification
from datetime import datetime
from multiprocessing import Process, Pool
from sensor_msgs.msg import Image
from rospy.numpy_msg import numpy_msg
from scipy.misc import imresize, imsave
from sklearn import svm
from sklearn.cross_validation import StratifiedKFold, cross_val_score
from skimage.feature import hog
OVERFEAT_DIM = 231
KINECT_WIDTH = 640
KINECT_HEIGHT = 480
# initialize overfeat. 0 = fast net, 1 = accurate net.
overfeat.init("../data/weights/net_weight_0", 0)
# list of unique classes as strings.
# indices correspond to the 'target' of each 'classification'.
classes = []
# list of training samples.
training = []
# each Sample is features data and the target of a classification.
Sample = namedtuple('Sample', ['features','depth','classification','time','image'])
# will train svm classifier on features extracted using overfeat
svm_clf = svm.LinearSVC()
# use `image` and `depth` in other functions to access kinect frames.
testing_images.append(index_mat[i][0, j][0])
#net_types = ["OverFeat", "GoogLeNet", "CaffeNet", "Cifar10"]
net_types = ["GoogLeNet", "CaffeNet", "Cifar10"]
num_images = [10, 50, 100, 500, 1000]
timing_data = {}
for net_type in net_types:
timing_data[net_type] = {}
for num_image in num_images:
print("Timing %s for %s images" % (net_type, num_image))
# OverFeat does not use caffe
if net_type == 'OverFeat':
overfeat.init(overfeat_root + 'data/default/net_weight_0', 0)
start = time.clock()
for i in range(num_image):
image = smush_overfeat_images(training_images[i])
b = overfeat.fprop(image)
end = time.clock()
elapsed = end - start
print(elapsed)
timing_data[net_type][num_image] = elapsed
# Use caffe for all other models
else:
def fit(self, X=None, y=None):
import overfeat # soft dep
overfeat.init(self.pretrained_params, self.network_size)
return self
from PIL import Image
import numpy
import skimage
from skimage.feature import daisy, hog
import subprocess
import overfeat
import numpy
from scipy.ndimage import imread
from scipy.misc import imresize
import database_file
import time
from find_obj import init_feature
from asift import affine_detect
print "Initializing overfeat"
overfeat.init('overfeat/overfeat/data/default/net_weight_0', 0)
#overfeat_initialized = False
def extract_overfeat(image_path):
# if overfeat_initialized == None or not overfeat_initialized:
# overfeat_initialized = True
print "Overfeat: ", image_path
image = imread(image_path)
print "Image shape: ", image.shape
if len(image.shape) == 2 or image.shape[2] == 2:
image = skimage.color.gray2rgb(image)
elif image.shape[2] == 4:
image_rgb = numpy.zeros((image.shape[0],image.shape[1], 3), numpy.uint8)
image_rgb[:,:,0] = image[:,:,0]