print("\n SAVED THE WEIGHTS AND MODEL !!!!")
def load_model_weights(name):
from keras.models import model_from_json
json_file = open(name + '.json', 'r')
loaded_model_json = json_file.read()
json_file.close()
loaded_model = model_from_json(loaded_model_json)
loaded_model.load_weights(name + ".h5")
loaded_model.compile(loss='categorical_crossentropy',
optimizer='adam',
metrics=['accuracy'])
return loaded_model
if __name__ == '__main__':
from datacollect import load_train_data, shape, margin
X_train, Y_train = load_train_data()
model = load_model_weights('v2_nfs_ai')
for i in range(100):
model.fit({
'right_input': X_train[0],
'left_input': X_train[1]
},
Y_train,
batch_size=2000,
nb_epoch=10,
verbose=1)
save_model(model, "v2_nfs_ai")
"""
import numpy as np
import cv2
from time import time
from keras.models import Sequential, Model
from keras.layers import *
from keras.callbacks import TensorBoard
from model import load_model_weights
from datacollect import load_train_data, shape, margin
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
model = load_model_weights("v4_nfs_iteration2_ai")
layer_dict = dict([(layer.name, layer) for layer in model.layers])
X_train, Y_train = load_train_data('third_batch')
model.summary()
test = Sequential()
test.add(layer_dict['right_input'])
test.add(layer_dict['conv2d_1'])
test.add(layer_dict['conv2d_2'])
#test.add(layer_dict['conv2d_3'])
#test.add(layer_dict['conv2d_4'])
#test.add(layer_dict['conv2d_5'])
side = 1
res = test.predict(X_train[side, :, :, :, :])
for i in range(X_train.shape[1]):
beta_2=0.999,
epsilon=1e-6,
decay=1e-6,
amsgrad=False)
loaded_model.compile(loss='categorical_crossentropy',
optimizer=adam,
metrics=['accuracy'])
return loaded_model
if __name__ == '__main__':
from datacollect import load_train_data, shape, margin
batches = ['first_batch', 'second_batch']
X_train, Y_train = load_train_data(batches[0])
for i in range(1, len(batches)):
x, y = load_train_data(batches[i])
print(x.shape, X_train.shape)
X_train = np.concatenate((X_train, x), axis=1)
Y_train = np.concatenate((Y_train, y), axis=0)
print("log: loaded data")
print("data size :", X_train.shape)
learning_rate = 0.001
#model = bifocal_nvidia(shape , margin , learning_rate)
#print("log: Model created")
model = load_model_weights('v3_nfs_ai', learning_rate)
X = X_train
Y = Y_train
@author: RAJAT
"""
import numpy as np
from time import time
from keras.models import Sequential, Model
from keras.layers import *
from keras.callbacks import TensorBoard
from model import load_model_weights
from datacollect import load_train_data, shape, margin
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
model = load_model_weights("v3_nfs_ai")
layer_dict = dict([(layer.name, layer) for layer in model.layers])
X_train, Y_train = load_train_data('second_batch')
model.summary()
test = Sequential()
test.add(layer_dict['right_input'])
test.add(layer_dict['conv2d_1'])
test.add(layer_dict['conv2d_2'])
#test.add(layer_dict['conv2d_3'])
#test.add(layer_dict['conv2d_4'])
#test.add(layer_dict['conv2d_5'])
img_no = 8000
side = 0
plt.imshow(X_train[side, img_no, :, :, 0], cmap='gray')
res = test.predict(X_train[side, :, :, :, :])