def load_piropo():
# img_rows, img_cols = 128, 128
ilist = set_para.lookfor(keyword = 'omni_1A', imagelist = None)
ilist = set_para.lookfor(keyword = 'omni_1B', imagelist = ilist)
ilist = set_para.lookfor(keyword = 'omni_2A', imagelist = ilist)
ilist = set_para.lookfor(keyword = 'omni_3A', imagelist = ilist)
frame = pandas.DataFrame.from_dict(ilist)
train, test = train_test_split(frame, random_state = 1)
# train = pandas.DataFrame.as_matrix(train)
# test = pandas.DataFrame.as_matrix(test)
Y_train = np.asarray(train.label.tolist())
Y_test = np.asarray(test.label.tolist())
X_train = np.asarray(train.picarray.tolist())
X_test = np.asarray(test.picarray.tolist())
X_train, Y_train, X_test, Y_test = set_para.preprocessing(X_train = X_train, y_train = Y_train, X_test = X_test, y_test = Y_test
, nb_classes = nb_classes, to_categorical = True)
# x = []
# y = []
# for item in ilist:
# x.append(item['picarray'])
# y.append(item['label'])
# x = np.asarray(x)
# y = np.asarray(y)
# y = np.reshape(y, (len(y),1))
# evalu.savexy(x,y, modelname, modelpath)
# x, y = evalu.loadxy(modelname, modelpath)
# X_train, Y_train, X_test, Y_test = set_para.preprocessing(x = x, y = y, split = True, nb_classes = nb_classes,
# to_categorical = True)
evalu.savexy(X_train, Y_train, 'piropo_train', modelpath)
evalu.savexy(X_test, Y_test, 'piropo_test', modelpath)
evalu.savexy(train, test, 'dataframe', modelpath)
return X_train, Y_train, X_test, Y_test, train, test
def load_stereo():
img_rows, img_cols = 36,18
x, y = set_para.readimage(label = 1, cutsize = False, resize = False,
path = '/home/workstation/Documents/humandataset/pedestrain_benchmarks/Stereo/TrainingData/Pedestrians/18x36/')
x, y = set_para.readimage(label = 0, cutsize = True, resize = False,
x = x, labelist = y, reh = 36, rew = 18,
path = '/home/workstation/Documents/humandataset/pedestrain_benchmarks/Stereo/TrainingData/NonPedestrians/')
X_train, Y_train, X_test, Y_test = set_para.preprocessing(x = x, y = y, split = True, nb_classes = nb_classes,
to_categorical = False)
return X_train, Y_train, X_test, Y_test
def load_google():
img_rows, img_cols = 32, 64
x, y = set_para.readimage(label = 1, cutsize = False, resize = True, reh = img_rows, rew = img_cols,
path = '/home/workstation/Documents/humandataset/beforethermal/1/pedestrain/')
x, y = set_para.readimage(label = 1, cutsize = False, resize = True,
x = x, labelist = y, reh = img_rows, rew = img_cols,
path = '/home/workstation/Documents/humandataset/beforethermal/1/people_on_the_street/')
x, y = set_para.readimage(label = 1, cutsize = False, resize = True,
x = x, labelist = y, reh = img_rows, rew = img_cols,
path = '/home/workstation/Documents/humandataset/beforethermal/1/traffic_warden/')
x, y = set_para.readimage(label = 0, cutsize = False, resize = True,
x = x, labelist = y, reh = img_rows, rew = img_cols,
path = '/home/workstation/Documents/humandataset/beforethermal/0/empty_city_road/')
x, y = set_para.readimage(label = 0, cutsize = False, resize = True,
x = x, labelist = y, reh = img_rows, rew = img_cols,
path = '/home/workstation/Documents/humandataset/beforethermal/0/empty_street/')
x, y = set_para.readimage(label = 0, cutsize = False, resize = True,
x = x, labelist = y, reh = img_rows, rew = img_cols,
path = '/home/workstation/Documents/humandataset/beforethermal/0/trees_street_empty/')
X_train, Y_train, X_test, Y_test = set_para.preprocessing(x = x, y = y, split = True, nb_classes = nb_classes,
to_categorical = False)
return X_train, Y_train, X_test, Y_test
# import plot
import evalu
sys.setrecursionlimit(10000)
modelpath = "/home/workstation/Documents/Cifar10/best/"
modelname = raw_input("Ask me == ")
batch_size = 32
nb_classes = 10
nb_epoch = 2
data_augmentation = True
img_rows, img_cols = 32, 32
img_channels = 3
(X_train, y_train), (X_test, y_test) = cifar10.load_data()
# X_train, y_train, X_test, y_test = set_para.slicing(X_train, y_train, X_test, y_test)
X_train, Y_train, X_test, Y_test = set_para.preprocessing(X_train, y_train, X_test, y_test)
changed_list = ['dense']
print (changed_list[0] + ' is changed at ' + strftime("%Y-%m-%d %H:%M:%S", gmtime()))
ds = 938
acc = []
avg_acc = []
m_hist = []
model = Sequential()
model.add(Convolution2D(32, 5, 5, border_mode='same',
input_shape=(img_channels, img_rows, img_cols)))
model.add(Activation('relu'))
model.add(Convolution2D(32, 5, 5))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.12))
def load_beforethermal():
# Seems these are random google pics resized into (2446, 3, 64, 32) ones
m = pickle.load(open( '/home/workstation/Documents/humandataset/beforethermal/' + 'beforethermal_x' + '.p', 'rb'))
y = pickle.load(open( '/home/workstation/Documents/humandataset/beforethermal/' + 'beforethermal_y' + '.p', 'rb'))
X_train, Y_train, X_test, Y_test = set_para.preprocessing(x = x, y = y, split = True, nb_classes = nb_classes)
return X_train, Y_train, X_test, Y_test
modelname = raw_input("Ask me == ")
batch_size = 32
nb_classes = 2
nb_epoch = 10
data_augmentation = True
img_rows, img_cols = 32, 32
img_channels = 3
(X_train, y_train), (X_test, y_test) = cifar100.load_data()
y_train = np.where((y_train < 74), y_train, 0)
y_train = np.where((y_train > 69), 1, 0)
y_test = np.where((y_test < 74), y_test, 0)
y_test = np.where((y_test > 69), 1, 0)
# X_train, y_train, X_test, y_test = set_para.slicing(X_train, y_train, X_test, y_test)
X_train, Y_train, X_test, Y_test = set_para.preprocessing(
X_train, y_train, X_test, y_test, nb_classes=nb_classes)
acc = []
avg_acc = []
m_hist = []
model = Sequential()
model.add(
Convolution2D(32,
5,
5,
border_mode='same',
input_shape=(img_channels, img_rows, img_cols)))
model.add(Activation('relu'))
model.add(Convolution2D(32, 5, 5))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
batch_size = 32
nb_classes = 2
nb_epoch = 100
data_augmentation = True
img_rows, img_cols = 36,18
img_channels = 3
x, y = set_para.readimage(label = 1, cutsize = False, resize = False,
path = '/home/workstation/Documents/humandataset/pedestrain_benchmarks/Stereo/TrainingData/Pedestrians/18x36/')
x, y = set_para.readimage(label = 0, cutsize = True, resize = False,
x = x, labelist = y, reh = 36, rew = 18,
path = '/home/workstation/Documents/humandataset/pedestrain_benchmarks/Stereo/TrainingData/NonPedestrians/')
X_train, Y_train, X_test, Y_test = set_para.preprocessing(x = x, y = y, split = True, nb_classes = nb_classes,
to_categorical = False)
acc = []
avg_acc = []
m_hist = []
inputs = Input(shape = (img_channels, img_rows, img_cols))
cn1 = Convolution2D(32,5,5, border_mode = 'same')(inputs)
relu1 = Activation('relu')(cn1)
pool1 = MaxPooling2D(pool_size = (2,2))(relu1)
do1 = Dropout(0.2)(pool1)
cn2 = Convolution2D(64,3,3, border_mode = 'same')(do1)
relu2 = Activation('relu')(cn2)
pool2 = MaxPooling2D(pool_size = (2,2))(relu2)