import pdb
test_datagen = ImageDataGenerator(rescale=1. / 255)
test_datagen.config['center_crop_size'] = (224, 224)
test_datagen.set_pipeline([center_crop])
dgdx_val = test_datagen.flow_from_directory(
'/home/nancy/mvmt_vid_dataset/test/',
read_formats={'png'},
target_size=(int(256 * (224 / 192.0)), int(192 * (224 / 192.0))),
batch_size=32,
shuffle=False,
class_mode=None)
test_datagen.fit_generator(dgdx_val, nb_iter=100)
validation_generator = dgdx_val
base_model = VGG16(input_tensor=(Input(shape=(224, 224, 3))),
weights='imagenet',
include_top=False)
#base_model = VGG16(input_tensor=(Input(shape=(224, 224, 3))), include_top=False)
#for layer in base_model.layers[:10]:
# layer.trainable = False
model = load_model("my_model.h5")
results = model.predict_generator(validation_generator, 32 * 5)
#pdb.set_trace()
dgdx = train_datagen.flow_from_directory(
'/home/wangnxr/dataset/vid_offset_0/train/',
read_formats={'png'},
target_size=(int(300), int(224)),
batch_size=32,
class_mode='binary')
dgdx_val = test_datagen.flow_from_directory(
'/home/wangnxr/dataset/vid_offset_0/test/',
read_formats={'png'},
target_size=(int(300), int(224)),
batch_size=32,
class_mode='binary')
train_datagen.fit_generator(dgdx, nb_iter=96)
test_datagen.fit_generator(dgdx_val, nb_iter=96)
train_generator=dgdx
validation_generator=dgdx_val
base_model = VGG16(input_tensor=(Input(shape=(3, 224, 224))), weights='imagenet', include_top=False)
#base_model = VGG16(input_tensor=(Input(shape=(224, 224, 3))), include_top=False)
x = base_model.output
x = Flatten(name='flatten')(x)
x = Dense(1024, name='fc1')(x)
x = BatchNormalization()(x)
x = Activation('relu')(x)
x = Dropout(0.5)(x)
x = Dense(256, name='fc2')(x)
