In [1]:

    

import os
import json
import numpy as np
import pandas as pd
import cv2
from sklearn.utils import shuffle
import tensorflow as tf

from keras.models import Model, Sequential
from keras.layers.core import Dense, Activation, Flatten, Dropout
from keras.layers.convolutional import Convolution2D, Cropping2D
from keras.layers.pooling import MaxPooling2D
from keras.layers import Lambda

import matplotlib.pyplot as plt
import matplotlib.image as mpimg
%matplotlib inline

import progressbar as pb


    

    




Using TensorFlow backend.

In [2]:

    

sample_data_file = "driving_log.csv"
sample_data = pd.read_csv(sample_data_file, sep = ",", skipinitialspace=True)

recover_data_file = "recover.csv"
recover_data = pd.read_csv(recover_data_file, sep = ",", skipinitialspace=True, header =None)
recover_data.columns = ["center","left","right","steering","throttle","brake","speed"]


    

In [3]:

    

recover_data["center"]=recover_data["center"].str.replace("/Applications/", "")
recover_data["left"]=recover_data["left"].str.replace("/Applications/", "")
recover_data["right"]=recover_data["right"].str.replace("/Applications/", "")
recover_data = recover_data[recover_data["steering"]!= 0.0]


    

In [4]:

    

print("number recover data:",len(recover_data))
print("number sample data from udacity: ",len(sample_data))


    

    




number recover data: 80
number sample data from udacity:  8036

In [5]:

    

sample_data = sample_data.append(recover_data)
print("together: ", len(sample_data))


    

    




together:  8116

In [6]:

    

angle_adjust = 0.21

center = sample_data[["center","steering"]]
left = sample_data[["left","steering"]]
right = sample_data[["right","steering"]]

left['steering'] = left['steering'] + angle_adjust
right['steering'] = right['steering'] - angle_adjust

center.columns=["path","steering"]
left.columns=["path","steering"]
right.columns=["path","steering"]


data = left.append(center).append(right)


    

    




/Users/Gon/anaconda3/envs/carnd-term1/lib/python3.5/site-packages/ipykernel/__main__.py:7: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy
/Users/Gon/anaconda3/envs/carnd-term1/lib/python3.5/site-packages/ipykernel/__main__.py:8: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy

In [16]:

    

steering = data["steering"].tolist()
data["steering_flip"] = -data['steering']
steering_flipped = data['steering_flip'].tolist()
#put together
steering_ready_to_draw = steering+steering_flipped


    

In [17]:

    

plt.hist(steering_ready_to_draw, 15, facecolor='blue')
plt.xlabel('steering angles')
plt.ylabel('number of data')
plt.title('distribution of steering angle from training data')
plt.axis([-1, 1, 0, 15000 ])
plt.grid(True)
plt.show()


    

In [26]:

    

##########
# pip install prograssbar2 for python 3.5
# http://stackoverflow.com/questions/13520622/script-to-show-progress
# http://pythonhosted.org/progressbar2/usage.html#bar-with-custom-widgets
#########

# progress = pb.ProgressBar(maxval = len(data)).start()
# progvar = 0


    

In [8]:

    

# path = data["path"].tolist()
# images = []
# images_flipped= []

# progress = pb.ProgressBar(maxval = len(data)).start()
# progvar = 0
# for img in path:
    
#     image =cv2.imread(img)
#     images.append(image)
#     image_flip = cv2.flip(image, 1)
#     images_flipped.append(image_flip)
    
#     progress.update(progvar)
#     progvar += 1
# print("finish loading images to array")
# #     print('path: ',img)
# #     plt.imshow(image)
# #     plt.imshow(image_flip)
# #     print('image', image)
# #     print('image flip: ', image_flip)
# #     print('images list: ', images)
# #     print('flip list: ', images_flipped)
    
    
# steering = data["steering"].tolist()

# print("done")


    

In [9]:

    

# #flip 
# data["steering_flip"] = -data['steering']
# steering_flipped = data['steering_flip'].tolist()
# #put together
# images_ready = images+images_flipped
# steering_ready = steering+steering_flipped

# print("all set")
# # X_train = np.array(images_ready)
# # Y_train = np.array(steering_ready)

# print('done')


    

In [10]:

    

def steering_model():
    model = Sequential()
    model.add(Lambda(lambda x: (x / 255.0) - 0.5, input_shape=(160,320,3)))
    model.add(Cropping2D(cropping=((70,20), (0,0))))
    model.add(Convolution2D(24,5,5,subsample=(2,2),activation = "relu"))
#     model.add(MaxPooling2D((2,2)))
    model.add(Convolution2D(32,5,5,subsample=(2,2),activation = "relu"))
    model.add(Convolution2D(48,5,5,subsample=(2,2),activation = "relu"))
    model.add(Convolution2D(64,3,3,activation = "relu"))
    model.add(Convolution2D(64,3,3,activation = "relu"))
    model.add(Flatten())
    model.add(Dropout(0.2))
    model.add(Dense(100, activation = "relu"))
    model.add(Dropout(0.3))
    model.add(Dense(50, activation = "relu"))
    model.add(Dropout(0.4))
    model.add(Dense(10, activation = "relu"))
    model.add(Dense(1))
#     model.summary()
    return model


    

In [11]:

    

def image_process(db):
    path = db["path"].tolist()
    images = []
    images_flipped= []

#     progress = pb.ProgressBar(maxval = len(db)).start()
#     progvar = 0
    
    for img in path:
        image =cv2.imread(img)
        images.append(image)
        image_flip = cv2.flip(image, 1)
        images_flipped.append(image_flip)

#         progress.update(progvar)
#         progvar += 1
#     print("finish loading images to array")
    
    steering = db["steering"].tolist()
    db["steering_flip"] = -db['steering']
    steering_flipped = db['steering_flip'].tolist()
    #put together
    images_ready = images+images_flipped
    steering_ready = steering+steering_flipped
    
    images_set, steering_set = shuffle(images_ready, steering_ready)
    return images_set, steering_set


    

In [12]:

    

def generator(x , y , batch_size):
    num_samples = len(x)
    while 1: # Loop forever so the generator never terminates
        shuffle(x, y)
        for offset in range(0, num_samples, batch_size):
#             batch_samples = samples[offset:offset+batch_size]

            images = x[offset:offset+batch_size]
            angles = y[offset:offset+batch_size]
            # trim image to only see section with road
            X_train = np.array(images)
            y_train = np.array(angles)
            yield shuffle(X_train, y_train)


    

In [13]:

    

def generator2(db, batch_size):
    
    num_samples = len(db)
    while 1: # Loop forever so the generator never terminates
#         shuffle(x, y)
        for offset in range(0, num_samples, batch_size):

            batch_db = db[offset:offset+batch_size]
            images, angles = image_process(batch_db)
            # trim image to only see section with road
            X_train = np.array(images)
            y_train = np.array(angles)
            yield shuffle(X_train, y_train)


    

In [14]:

    

epoch = 10
batch = 64

# x_train, y_train = image_process(data)
# x_test, y_test = image_process(center)
# train_generator = generator(x_train, y_train, batch)
# test_generator = generator(x_test, y_test, batch)

train_generator = generator2(data, batch)
test_generator = generator2(center, batch)

model = steering_model()
model.compile(optimizer="adam", loss = "mse")
history = model.fit_generator(train_generator, samples_per_epoch= len(data)*2, validation_data=test_generator, 
                    nb_val_samples=len(center), nb_epoch=epoch)
# model.fit(X_train,Y_train, validation_split = 0.2, shuffle = True, nb_epoch=epoch)
model.save('model2.h5')

print("model trained")


json_str = model.to_json()
with open('model2.json','w') as f:
    f.write(json_str)
print('json file saved')


    

    




Epoch 1/10






    




/Users/Gon/anaconda3/envs/carnd-term1/lib/python3.5/site-packages/ipykernel/__main__.py:20: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy






    




48696/48696 [==============================] - 1086s - loss: 0.0292 - val_loss: 0.0297
Epoch 2/10
48696/48696 [==============================] - 1103s - loss: 0.0208 - val_loss: 0.0268
Epoch 3/10
48696/48696 [==============================] - 1083s - loss: 0.0192 - val_loss: 0.0311
Epoch 4/10
48696/48696 [==============================] - 1016s - loss: 0.0181 - val_loss: 0.0256
Epoch 5/10
48696/48696 [==============================] - 1033s - loss: 0.0177 - val_loss: 0.0321
Epoch 6/10
48696/48696 [==============================] - 1013s - loss: 0.0179 - val_loss: 0.0228
Epoch 7/10
48696/48696 [==============================] - 1019s - loss: 0.0165 - val_loss: 0.0293
Epoch 8/10
48696/48696 [==============================] - 1027s - loss: 0.0153 - val_loss: 0.0268
Epoch 9/10
48696/48696 [==============================] - 1030s - loss: 0.0142 - val_loss: 0.0231
Epoch 10/10
48696/48696 [==============================] - 1039s - loss: 0.0141 - val_loss: 0.0271
model trained
json file saved

In [15]:

    

model.summary()


    

    




____________________________________________________________________________________________________
Layer (type)                     Output Shape          Param #     Connected to                     
====================================================================================================
lambda_1 (Lambda)                (None, 160, 320, 3)   0           lambda_input_1[0][0]             
____________________________________________________________________________________________________
cropping2d_1 (Cropping2D)        (None, 70, 320, 3)    0           lambda_1[0][0]                   
____________________________________________________________________________________________________
convolution2d_1 (Convolution2D)  (None, 33, 158, 24)   1824        cropping2d_1[0][0]               
____________________________________________________________________________________________________
convolution2d_2 (Convolution2D)  (None, 15, 77, 32)    19232       convolution2d_1[0][0]            
____________________________________________________________________________________________________
convolution2d_3 (Convolution2D)  (None, 6, 37, 48)     38448       convolution2d_2[0][0]            
____________________________________________________________________________________________________
convolution2d_4 (Convolution2D)  (None, 4, 35, 64)     27712       convolution2d_3[0][0]            
____________________________________________________________________________________________________
convolution2d_5 (Convolution2D)  (None, 2, 33, 64)     36928       convolution2d_4[0][0]            
____________________________________________________________________________________________________
flatten_1 (Flatten)              (None, 4224)          0           convolution2d_5[0][0]            
____________________________________________________________________________________________________
dropout_1 (Dropout)              (None, 4224)          0           flatten_1[0][0]                  
____________________________________________________________________________________________________
dense_1 (Dense)                  (None, 100)           422500      dropout_1[0][0]                  
____________________________________________________________________________________________________
dropout_2 (Dropout)              (None, 100)           0           dense_1[0][0]                    
____________________________________________________________________________________________________
dense_2 (Dense)                  (None, 50)            5050        dropout_2[0][0]                  
____________________________________________________________________________________________________
dropout_3 (Dropout)              (None, 50)            0           dense_2[0][0]                    
____________________________________________________________________________________________________
dense_3 (Dense)                  (None, 10)            510         dropout_3[0][0]                  
____________________________________________________________________________________________________
dense_4 (Dense)                  (None, 1)             11          dense_3[0][0]                    
====================================================================================================
Total params: 552,215
Trainable params: 552,215
Non-trainable params: 0
____________________________________________________________________________________________________

In [27]:

    

hist= history.history
training_set, validation_set = hist["loss"], hist["val_loss"]
plt.plot(training_set, label= "training_set")
plt.plot(validation_set, label = "validation_set")
plt.axis([0,10,0,0.05])
plt.legend()
plt.xlabel('number of epoch')
plt.ylabel('mse loss')
plt.title('model mean squared error loss')
plt.show()


    

In [ ]: