def main():
try:
# initialize variable
db = DataBase()
settings = Config.Settings
if args["feature"] is None:
feature = settings["FEATURE"]
else:
feature = args["feature"]
if args["type"] is None:
kmeansType = Config.KMeans["TYPE"]
else:
kmeansType = args["type"]
filename = "{}/{}_{}_kmeans{}_{}".format(
settings["ROOT_DATASET_FOLDER"], settings["TRAIN_DATASET"],
feature, kmeansType, settings["FEATURE_FILE"])
if db.checkFile(filename):
(image_paths, im_features, idf, numWords, voc) = db.read(filename)
else:
descriptorHandler = Descriptor(Config)
numWords = Config.KMeans["NUM_WORDS"]
iterations = Config.KMeans["ITER"]
imageFiles = glob.glob("{}/{}/*.*".format(
settings["ROOT_DATASET_FOLDER"], settings["TRAIN_DATASET"]))
if len(imageFiles) > settings["MAX_FILES"]:
np.random.shuffle(imageFiles)
imageFiles = imageFiles[:settings["MAX_FILES"]]
# Extract features of images
(image_paths, des_list,
descriptors) = extractFeatures(descriptorHandler, feature,
imageFiles)
# Create descriptors
# descriptors = createDescriptors(des_list, imageFiles);
# Perform k-means clustering
if (descriptors.shape[0] / numWords) < 100:
numWords = numWords / 100
(voc, variance) = kMeansClustering(kmeansType, descriptors,
numWords, iterations)
# Calculate the TF-IDF of features
filename = "{}/{}_{}_kmeans{}_{}".format(
settings["ROOT_DATASET_FOLDER"], settings["TRAIN_DATASET"],
feature, kmeansType, settings["FEATURE_FILE"])
(im_features, idf) = calculateTFIDF(des_list, voc, variance,
numWords, len(image_paths))
db.write(filename, (image_paths, im_features, idf, numWords, voc))
except Exception as e:
tb = traceback.format_exc()
print 'Error found: %s' % (tb)
from database import DataBase
from config import Config
from descriptors.feature import Descriptor
# Load the classifier, class names, scaler, number of clusters and vocabulary (voc)
db = DataBase()
descriptorHandler = Descriptor(Config)
settings = Config.Settings
Datasets = {}
for key, value in Config.Features.iteritems():
filename = "{}/{}_{}_kmeans{}_{}".format(settings["ROOT_DATASET_FOLDER"],
settings["TRAIN_DATASET"], value,
Config.KMeans["TYPE"],
settings["FEATURE_FILE"])
if db.checkFile(filename):
(image_paths, im_features, idf, numWords, voc) = db.read(filename)
Datasets[value] = (image_paths, im_features, idf, numWords, voc)
def getQueryImage(image_path, cx=0, cy=0, cw=0, ch=0):
if image_path.find("http") != -1:
imgdata = urllib.urlopen(image_path)
arr = np.asarray(bytearray(imgdata.read()), dtype=np.uint8)
query = cv2.imdecode(arr, -1)
else:
if (image_path.find("/") == -1) and (image_path.find("\\") == -1):
query = cv2.imread("{}/{}".format(
Config.Resources["UPLOAD_FOLDER"], image_path))
elif (image_path.find(settings["ROOT_DATASET_FOLDER"]) == -1):
query = cv2.imread("{}/{}".format(settings["ROOT_DATASET_FOLDER"],
