def feature_extractor():
#Abrir arquivo que vai receber as features dos arquivos de imagens
file = "image_features_train_20_15.arff"
text_file = open(file, "a")
trainPath = "C:/Python27/ARP/Dataset/train2"
dirList = os.listdir(trainPath)
edge = EdgeHistogramFeatureExtractor(20)
for dirName in dirList:
fileList = os.listdir(trainPath + '/' + dirName)
maxsize = len(fileList)
for i in range(0, maxsize):
#print fileList[i]
filePATH = (trainPath + '/' + dirName + '/' + fileList[i])
img = Image(filePATH)
#Feature - Edge histogram
edge_vecs = edge.extract(img)
edge_fields = edge.getFieldNames()
for y in range(0, edge.getNumFields()):
#print edge_fields[y], '=', edge_vecs[y]
text_file.write(str(edge_vecs[y]) + ",")
#Feature - Hue
hue = HueHistogramFeatureExtractor(15)
hue_vecs = hue.extract(img)
hue_fields = hue.getFieldNames()
for i in range(0, hue.getNumFields()):
text_file.write(str(hue_vecs[i]) + ",")
text_file.write(dirName + "\n")
text_file.close()
def hue():
hue = HueHistogramFeatureExtractor()
a = np.array(hue.extract(watchman.imgbank[0]))
if len(watchman.imgbank) > 1:
b = np.array(hue.extract(watchman.imgbank[1]))
else:
b = np.array(hue.extract(watchman.imgbank[0]))
AandB = np.sum(np.square(a-b))
return AandB
def hue():
hue = HueHistogramFeatureExtractor()
a = np.array(hue.extract(watchman.imgbank[0]))
if len(watchman.imgbank) > 1:
b = np.array(hue.extract(watchman.imgbank[1]))
else:
b = np.array(hue.extract(watchman.imgbank[0]))
AandB = np.sum(np.square(a - b))
return AandB
def getFeatureExtractors(self):
hueHistogramFeatureExtractor = HueHistogramFeatureExtractor(10)
edgeHistogramFeatureExtractor = EdgeHistogramFeatureExtractor(10)
haarLikeFeatureExtractor = HaarLikeFeatureExtractor(
fname='/Users/burton/Downloads/SimpleCV/SimpleCV/Features/haar.txt'
)
return [
hueHistogramFeatureExtractor, edgeHistogramFeatureExtractor,
haarLikeFeatureExtractor
]
def createExtractor(self, extractorName, trainPaths=[]):
if (extractorName == 'hue'):
extractor = HueHistogramFeatureExtractor(10)
elif (extractorName == 'edge'):
extractor = EdgeHistogramFeatureExtractor(10)
elif (extractorName == 'haar'):
extractor = HaarLikeFeatureExtractor(fname='haar.txt')
elif (extractorName == 'bof'):
extractor = BOFFeatureExtractor()
extractor.generate(trainPaths, imgs_per_dir=40)
# need to build the vocabuary (feature words) for bag of feature
# extractor.generate(trainPaths, imgs_per_dir=40)
return extractor
def createExtractor(self, extractorName, trainPaths=[]):
if (extractorName == 'hue'):
extractor = HueHistogramFeatureExtractor(10)
elif (extractorName == 'edge'):
extractor = EdgeHistogramFeatureExtractor(10)
elif (extractorName == 'haar'):
extractor = HaarLikeFeatureExtractor(fname='haar.txt')
elif (extractorName == 'bof'):
extractor = BOFFeatureExtractor()
extractor.generate(trainPaths, imgs_per_dir=40)
# need to build the vocabuary (feature words) for bag of feature
# extractor.generate(trainPaths, imgs_per_dir=40)
return extractor
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler
from sklearn.ensemble import AdaBoostClassifier
from sklearn import cross_validation
from SimpleCV import Image
from SimpleCV import HueHistogramFeatureExtractor
from SimpleCV import HaarLikeFeatureExtractor
from SimpleCV import EdgeHistogramFeatureExtractor
import numpy as np
from glob import glob
import pickle
# model to extract feature
hhfe = HueHistogramFeatureExtractor(10)
ehfe = EdgeHistogramFeatureExtractor(10)
haarfe = HaarLikeFeatureExtractor('haar.txt')
# Give path of the training folder
images = glob('./fruits/*')
# Extract features and target labels for training
def get_feature_labels():
features = list()
labels = list()
for im in images:
try:
img = Image(im)
labels.append(im[:-2])