x = Dense(10, init=init)(x)
x = ELU()(x)
out = Dense(1, init=init)(x)
model = Model(input=input_frame, output=out)
if summary:
model.summary()
return model
if __name__ == '__main__':
# split udacity csv data into training and validation
train_data, val_data = split_train_val(csv_driving_data='data/driving_log.csv')
# get network model and compile it (default Adam opt)
nvidia_net = get_nvidia_model(summary=True)
nvidia_net.compile(optimizer='adam', loss='mse')
# json dump of model architecture
with open('logs/model.json', 'w') as f:
f.write(nvidia_net.to_json())
# define callbacks to save history and weights
checkpointer = ModelCheckpoint('checkpoints/weights.{epoch:02d}-{val_loss:.3f}.hdf5')
logger = CSVLogger(filename='logs/history.csv')
# start the training
nvidia_net.fit_generator(generator=generate_data_batch(train_data, augment_data=True, bias=CONFIG['bias']),
# Load data
df_vehicles1 = load_data('object-detection-crowdai', 'labels.csv', verbose=False)
df_vehicles2 = load_data('object-dataset', 'labels.csv', verbose=False)
# Concatenate
df_vehicles = pd.concat([df_vehicles1,df_vehicles2]).reset_index()
df_vehicles = df_vehicles.drop('index', 1)
df_vehicles.columns =['File_Path','Frame','Label','ymin','xmin','ymax','xmax']
print(len(df_vehicles))
# split udacity csv data into training and validation
train_data, val_data = split_train_val(df_vehicles)
### Generator
model = get_unet()
model.summary()
if config.mode == 'train':
# training_gen = generate_data_batch(train_data, df_vehicles, config.batch_size)
training_gen = generate_train_batch(train_data, config.batch_size)
eval_gen = generate_train_batch(val_data, config.batch_size)
smooth = 1.
x = Dense(10, init=init)(x)
x = ELU()(x)
out = Dense(1, init=init)(x)
model = Model(input=input_frame, output=out)
if summary:
model.summary()
return model
if __name__ == '__main__':
# split udacity csv data into training and validation
train_data, val_data = split_train_val(csv_driving_data='data/driving_log.csv')
# get network model and compile it (default Adam opt)
nvidia_net = get_nvidia_model(summary=True)
nvidia_net.compile(optimizer='adam', loss='mse')
# json dump of model architecture
with open('logs/model.json', 'w') as f:
f.write(nvidia_net.to_json())
# define callbacks to save history and weights
checkpointer = ModelCheckpoint('checkpoints/weights.{epoch:02d}-{val_loss:.3f}.hdf5')
logger = CSVLogger(filename='logs/history.csv')
# start the training
nvidia_net.fit_generator(generator=generate_data_batch(train_data, augment_data=True, bias=CONFIG['bias']),
x = Dropout(0.2)(x)
x = Conv2D(48, (5, 5), strides=(2, 2))(x)
x = ELU()(x)
x = Dropout(0.2)(x)
x = Conv2D(64, (3, 3))(x)
x = ELU()(x)
x = Dropout(0.2)(x)
x = Conv2D(64, (3, 3))(x)
x = ELU()(x)
x = Dropout(0.2)(x)
x = Flatten()(x)
x = Dense(100)(x)
x = ELU()(x)
x = Dense(10)(x)
x = ELU()(x)
out = Dense(1)(x)
model = Model(inputs=input_frame, outputs=out)
model.compile(optimizer='adam', loss='mse')
model.summary()
return model
if __name__ == '__main__':
train, test = split_train_val('/Users/dwang/self-driving-car/project_3_behavioral_cloning/data/driving_log.csv')
model = get_model()
model.fit_generator(generator=generate_data_batch(train, augment_data=True, bias=BIAS),
steps_per_epoch=BATCH_SIZE,
epochs=50,
validation_data=generate_data_batch(test, augment_data=False, bias=1.0),
validation_steps=BATCH_SIZE*100)