def pre_trained_model(left_im, right_im, reuse=False):
# Resnet model tensors have changed. Not compatible with checkpoint tensors.
'''
from tensorflow.contrib.slim.nets import resnet_v1
import tensorflow.contrib.slim as slim
# Create graph
inputs = tf.placeholder(tf.float32, shape=[64, input.shape[1].value, \
input.shape[2].value, input.shape[3].value])
with slim.arg_scope(resnet_v1.resnet_arg_scope()):
net, end_points = resnet_v1.resnet_v1_50(inputs, is_training=False)
saver = tf.train.Saver()
with tf.Session() as sess:
pdb.set_trace()
saver.restore(sess, tf.train.latest_checkpoint('resnet50_ckpt'))
representation_tensor = sess.graph.get_tensor_by_name('feature_generator/resnet_v1_50/pool5:0')
#img = ... #load image here with size [1, 224,224, 3]
#pdb.set_trace()
features = sess.run(representation_tensor, {'Placeholder:0': input})
'''
import tensornets as nets
# tf.disable_v2_behavior() # for TF 2
left_feats = nets.ResNet50(left_im, is_training=True, reuse=tf.AUTO_REUSE)
left_feats.pretrained()
right_feats = nets.ResNet50(right_im,
is_training=True,
reuse=tf.AUTO_REUSE)
right_feats.pretrained()
merged_features = tf.abs(tf.subtract(left_feats, right_feats))
return merged_features, left_feats, right_feats
def train_resnet50(
image_data_loader: ImageDatasetLoader,
parameter: ResNet50TrainParameter,
):
classes = len(image_data_loader.image_classes)
dataset = image_data_loader.build_dataset()
inputs = tf.placeholder(tf.float32, [None, 224, 224, 3])
outputs = tf.placeholder(tf.float32, [None, classes])
model = nets.ResNet50(inputs, is_training=True, classes=classes)
loss = tf.losses.softmax_cross_entropy(outputs, model)
train = tf.train.GradientDescentOptimizer(parameter.learning_rate) \
.minimize(loss)
saver = tf.train.Saver()
it = dataset.make_one_shot_iterator()
get_next = it.get_next()
with tf.Session() as sess:
sess.run(model.pretrained())
while True:
try:
image_data, label = sess.run(get_next)
except tf.errors.OutOfRangeError:
break
sess.run(train, {inputs: image_data, outputs: label})
save_path = saver.save(sess, parameter.checkpoint_save_path)
print('model checkpoint saved to: {}'.format(save_path))
featlayer = model.get_outputs()[-4] # ?x4032
bs = 250
elif modelname == 'VGG19':
model = nets.VGG19(inputs)
featlayer = model.get_outputs()[-4] # ?x4096
bs = 100
elif modelname == 'MobileNet25':
model = nets.MobileNet25(inputs)
featlayer = model.get_outputs()[-4]
bs = 2500
elif modelname == 'SqueezeNet':
model = nets.SqueezeNet(inputs)
featlayer = model.get_outputs()[-2]
bs = 1000
elif modelname == 'ResNet50':
model = nets.ResNet50(inputs)
featlayer = model.get_outputs()[-3] # 'avgpool:0', ?x2048
bs = 500
elif modelname == 'InceptionResNet2':
model = nets.InceptionResNet2(inputs)
featlayer = model.get_outputs()[-4]
bs = 250
else:
print('Unknown model: ', modelname)
raise SystemExit
model_pretrained = model.pretrained()
# ind selects which animal to extract features for
try:
ind = int(sys.argv[2])
def do_extract(args):
if args.seed is not None:
np.random.seed(args.seed)
FILENAMES = sorted(glob.glob(args.images))
nimages = len(FILENAMES)
if nimages == 0:
print("Did not find any images in '{}".format(args.images))
return
if args.verbose:
print("Processing {} images".format(nimages))
if args.model == 'NASNetAlarge':
target_size, crop_size = 331, 331
else:
target_size, crop_size = 256, 224
inputs = tf.placeholder(tf.float32, [None, crop_size, crop_size, 3])
if args.model == 'NASNetAlarge':
model = nets.NASNetAlarge(inputs)
bs = 250
elif args.model == 'VGG19':
model = nets.VGG19(inputs)
bs = 200
elif args.model == 'MobileNet25':
model = nets.MobileNet25(inputs)
bs = 2500
elif args.model == 'SqueezeNet':
model = nets.SqueezeNet(inputs)
bs = 1000
elif args.model == 'ResNet50':
model = nets.ResNet50(inputs)
bs = 500
else:
raise ValueError # this should not happen
model_pretrained = model.pretrained()
if args.batchsize:
bs = args.batchsize # overwrite default batchsize
nchunks = (nimages+bs-1)//bs
PREDS = [] # beware: we store all activations (#images x #classes x sizeof(float))
with tf.Session() as sess:
sess.run(model_pretrained)
for i in xrange(nchunks):
images = []
for filename in FILENAMES[i*bs:(i+1)*bs]:
img = utils.load_img(filename, target_size=target_size, crop_size=crop_size)
if args.blur:
img = apply_blur(img, args.blur)
if args.noise:
img = apply_noise(img, args.noise)
if args.dead:
img = apply_deadpixels(img, args.dead)
if args.dark:
img = apply_scaletoblack(img, args.dark)
if args.bright:
img = apply_scaletoblack(img, args.bright)
if args.geometry:
img = apply_geometry(img, args.geometry)
images.append( img.squeeze() )
images = model.preprocess(np.asarray(images))
preds = sess.run(model, {inputs: images})
PREDS.extend(preds)
if args.verbose:
print('Processed chunk {} of {}'.format(i, nchunks))
print('Most recent prediction:', utils.decode_predictions(preds, top=1)[0])
PREDS = np.asarray(PREDS)
if args.output:
np.savetxt(args.output, PREDS.max(axis=1), fmt='%.12f')
if args.rawoutput:
np.savez_compressed(args.rawoutput, PREDS)
if args.labeloutput:
np.savetxt(args.labeloutput, PREDS.argmax(axis=1), fmt='%d')
if args.verbose:
print("Done.")
if __name__ == '__main__':
# processing cmd line args
parser = argparse.ArgumentParser('generate layer projections')
g1 = parser.add_argument_group('computation options')
g1.add_argument('--layer_no', type=int, default=0)
g1.add_argument('--batchsize', type=int, default=12)
args = parser.parse_args()
# Filter datasets
(Xtrain, Ytrain), (Xtest, Ytest) = tf.keras.datasets.cifar10.load_data()
Xtrain, Ytrain = filter_photo(Xtrain, Ytrain)
max_data = Xtrain.shape[0]
# Configure
inputs = tf.placeholder(tf.float32, [None, 224, 224, 3])
model = nets.ResNet50(inputs, is_training=False)
load_model_weights = model.pretrained()
dir_str = 'pickle_data/layer_{layer_no}'.format(layer_no=args.layer_no)
directory = os.makedirs(dir_str, exist_ok=True)
# Filter layers
tensor_names = []
with open('pickle_data/tensor_names.pickle', 'rb') as ph:
tensor_names = [
p for p, name in enumerate(pickle.load(ph)) if 'relu' in name
]
if args.layer_no is 0:
print("Batch Size: ", args.batchsize)
print("Max Data: ", max_data)