if not os.path.exists(input_dir) and not os.path.isdir(input_dir):

print("Cannot access feature files directory: {}!".format(input_dir))

sys.exit(-1)

train_set = {}

test_set = {}

directory_list = glob.glob(input_dir + '/' + file_pattern)

for featured_file in directory_list:

file_name = os.path.basename(featured_file)

key = file_name[0:re.search("\d", file_name).start() - 1]

if train_set.get(key) is None:

train_set[key] = []

if len(train_set[key]) >= number_for_train:

if test_set.get(key) is None:

test_set[key] = []

test_set[key].append(file_name)

else:

train_set[key].append(file_name)

if os.path.exists('./features/train_features.txt'):

os.remove('./features/train_features.txt')

if os.path.exists('./features/test_features.txt'):

os.remove('./features/test_features.txt')

for file_name in train_set['yorkshire_terrier']:

with open(input_dir + '/' + file_name, 'r') as d:

data = json.load(d)

data = '0,' + ",".join(map(str, data))

with open('./features/train_features.txt', 'a') as f:

print(data, file=f)

for file_name in train_set['wheaten_terrier']:

with open(input_dir + '/' + file_name, 'r') as d:

data = json.load(d)

data = '1,' + ",".join(map(str, data))

with open('./features/train_features.txt', 'a') as f:

print(data, file=f)

for file_name in test_set['yorkshire_terrier']:

with open(input_dir + '/' + file_name, 'r') as d:

data = json.load(d)

data = '0,' + ",".join(map(str, data))

with open('./features/test_features.txt', 'a') as f:

print(data, file=f)

for file_name in test_set['wheaten_terrier']:

with open(input_dir + '/' + file_name, 'r') as d:

data = json.load(d)

data = '1,' + ",".join(map(str, data))

with open('./features/test_features.txt', 'a') as f:

# Load training data

data = sc.textFile('./features/train_features.txt')

parsed_data = data.map(make_labeled_point)

# Build the model

model = SVMWithSGD.train(parsed_data, iterations=100)

# Evaluate the model on training data

labels_and_preds = parsed_data.map(lambda p: (p.label, model.predict(p.features)))

train_err = labels_and_preds.filter(lambda lp: lp[0] != lp[1]).count() / float(parsed_data.count())

print("Training Error = " + str(train_err))

if os.path.exists('./model/ImageRecognitionModel') and os.path.isdir('./model/ImageRecognitionModel'):

shutil.rmtree('./model/ImageRecognitionModel')

# Save the model

i = 0

args = sys.argv[1:]

if len(args) != 4 and len(args) != 1:

usage(sys.argv[0])

sys.exit(-1)

input_dir = ""

size = 0

test_data = False

while i < len(args):

if args[i] == "--dir":

i += 1

input_dir = args[i]

if len(input_dir) == 0:

print("Input directory name is required")

sys.exit(-1)

elif args[i] == "--size":

i += 1

if not args[i].isdigit():

print("Train size by class ({}) must be a number".format(args[i]))

sys.exit(-1)

else:

size = int(args[i])

elif args[i] == "--test":

test_data = True

i += 1

if size > 0:

train_set, test_set = split_sets(input_dir, "*.json", size)

create_features(train_set, test_set, input_dir)

create_model()

if test_data:

# Load test data

data = sc.textFile('./features/test_features.txt')

parsed_data = data.map(make_labeled_point)

# Load the model

model = SVMModel.load(sc, './model/ImageRecognitionModel')

# Try the model against test data

labels_and_preds = parsed_data.map(lambda p: (p.label, model.predict(p.features)))

test_err = labels_and_preds.filter(lambda lp: lp[0] != lp[1]).count() / float(parsed_data.count())