def main(self):
with tf.Graph().as_default():
self.processed_tensor_list = map(
self.image_preprocessor, self.files)
# Iterate over a map object
for tensor_tuple in self.processed_tensor_list:
# Create the model, use the default arg scope to configure the
# batch norm parameters.
with self.slim.arg_scope(inception3.inception_v3_arg_scope()):
logits, _ = inception3.inception_v3(
tensor_tuple[1], num_classes=1001, is_training=False)
# Append a tuple (image, probability)
self.image_and_probabilities.append(
(tensor_tuple[0], tf.nn.softmax(logits)))
self.init_fn = self.slim.assign_from_checkpoint_fn(
os.path.join(self.checkpoints_dir, 'inception_v3.ckpt'),
self.slim.get_model_variables('InceptionV3'))
with tf.Session() as sess:
self.init_fn(sess)
for idx in range(self.dim):
print('Classifying on image' + str(idx))
_, probabilities = sess.run([self.image_and_probabilities[idx][
0], self.image_and_probabilities[idx][1]])
probabilities = probabilities[0, 0:]
sorted_inds = [i[0] for i in sorted(
enumerate(-probabilities), key=lambda x:x[1])]
names = imagenet.create_readable_names_for_imagenet_labels()
temp_array = []
for i in range(self.top_k):
index = sorted_inds[i]
temp_array.append(names[index])
if self.print_flag:
print('Probability %0.2f%% => [%s]' % (
probabilities[index], names[index]))
if any(self.keyword in s for s in temp_array):
self.accuracy += 1
print('Classification Accuracy ====> ' +
str(tf.divide(self.accuracy, self.dim)))
if self.save_result_to_file:
with open('Inception_v3_Results.txt', 'wb') as f:
f.write('Classification Accuracy\n')
f.write(str(tf.divide(self.accuracy, self.dim)))
def get_model(cfg, pretrained=False, load_param_from_ours=False):
if load_param_from_ours:
pretrained = False
model = None
num_classes = cfg.num_classes
if cfg.model == 'custom':
from models import custom_net
if cfg.patch_size == 64:
model = custom_net.net_64(num_classes = num_classes)
elif cfg.patch_size == 32:
model = custom_net.net_32(num_classes = num_classes)
else:
print('Do not support present patch size %s'%cfg.patch_size)
#model = model
elif cfg.model == 'googlenet':
from models import inception_v3
model = inception_v3.inception_v3(pretrained = pretrained, num_classes = num_classes)
elif cfg.model == 'vgg':
from models import vgg
if cfg.model_info == 19:
model = vgg.vgg19_bn(pretrained = pretrained, num_classes = num_classes)
elif cfg.model_info == 16:
model = vgg.vgg16_bn(pretrained = pretrained, num_classes = num_classes)
elif cfg.model == 'resnet':
from models import resnet
if cfg.model_info == 18:
model = resnet.resnet18(pretrained= pretrained, num_classes = num_classes)
elif cfg.model_info == 34:
model = resnet.resnet34(pretrained= pretrained, num_classes = num_classes)
elif cfg.model_info == 50:
model = resnet.resnet50(pretrained= pretrained, num_classes = num_classes)
elif cfg.model_info == 101:
model = resnet.resnet101(pretrained= pretrained, num_classes = num_classes)
if model is None:
print('not support :' + cfg.model)
sys.exit(-1)
if load_param_from_ours:
print('loading pretrained model from {0}'.format(cfg.init_model_file))
checkpoint = torch.load(cfg.init_model_file)
model.load_state_dict(checkpoint['model_param'])
model.cuda()
print('shift model to parallel!')
model = torch.nn.DataParallel(model, device_ids=cfg.gpu_id)
return model
def get_model(cfg, pretrained=True, load_param_from_folder=False):
if load_param_from_folder:
pretrained = False
model = None
num_classes = cfg.num_classes
if cfg.model == 'googlenet':
from models import inception_v3
model = inception_v3.inception_v3(pretrained = pretrained, num_classes = num_classes)
elif cfg.model == 'vgg':
from models import vgg
if cfg.model_info == 19:
model = vgg.vgg19_bn(pretrained = pretrained, num_classes = num_classes)
elif cfg.model_info == 16:
model = vgg.vgg16_bn(pretrained = pretrained, num_classes = num_classes)
elif cfg.model == 'resnet':
from models import resnet
if cfg.model_info == 18:
model = resnet.resnet18(pretrained= pretrained, num_classes = num_classes)
elif cfg.model_info == 34:
model = resnet.resnet34(pretrained= pretrained, num_classes = num_classes)
elif cfg.model_info == 50:
model = resnet.resnet50(pretrained= pretrained, num_classes = num_classes)
elif cfg.model_info == 101:
model = resnet.resnet101(pretrained= pretrained, num_classes = num_classes)
if model is None:
print('not support :' + cfg.model)
sys.exit(-1)
if load_param_from_folder:
print('loading pretrained model from {0}'.format(cfg.init_model_file))
checkpoint = torch.load(cfg.init_model_file)
model.load_state_dict(checkpoint['model_param'])
print('shift model to parallel!')
model = torch.nn.DataParallel(model, device_ids=cfg.gpu_id)
return model
classes = ['normal', 'monotone', 'screenshot', 'unknown']
config = {
'batch_size':
64,
'train_split':
0.8,
'lr':
0.000005,
'n_epochs':
100,
'print_every':
1,
'cuda':
True if cuda.is_available() else False,
'model':
inception_v3(pretrained=True, freeze=True),
'criterion':
nn.NLLLoss(),
'transform':
transforms.Compose([
transforms.Resize(size=299),
# transforms.RandomRotation(degrees=15),
# transforms.ColorJitter(),
transforms.RandomHorizontalFlip(),
# transforms.RandomResizedCrop(size=299),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
]),
'early_stop':
100,
'balance':
graph = tf.Graph() # no necessiry
with graph.as_default():
# 1. Construct a graph representing the model.
x = tf.placeholder(tf.float32, [None, 784]) # Placeholder for input.
y = tf.placeholder(tf.float32, [None, 10]) # Placeholder for labels.
NN = 1 # type of neural network
if NN == 1: # SLIM inception
x_images = tf.reshape(x, [-1, 28, 28, 1])
x_images = tf.image.resize_images(x_images, [299, 299])
logits, end_points = inception.inception_v3(x_images,
num_classes=10,
is_training=True)
if NN == 2: # SLIM simple NN
print('use slim')
def lenet(images):
net = slim.conv2d(images, 20, [5, 5], scope='conv1')
net = slim.max_pool2d(net, [2, 2], scope='pool1')
net = slim.conv2d(net, 50, [5, 5], scope='conv2')
net = slim.max_pool2d(net, [2, 2], scope='pool2')
net = slim.flatten(net, scope='flatten3')
net = slim.fully_connected(net, 500, scope='fc4')
net = slim.fully_connected(net,
10,