In [1]:
%matplotlib inline
import cv2
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
from sklearn.preprocessing import normalize
In [2]:
df1 = pd.read_csv('../data/raw/sample/driving_log.csv')
In [3]:
columns = df1.columns
In [4]:
df1_count = len(df1)
df1_count
Out[4]:
In [5]:
df1.info()
In [6]:
df1.head()
Out[6]:
In [7]:
df1_summary = df1.describe().transpose()
df1_summary_cols = df1_summary.index
print(df1_summary_cols)
df1_summary
Out[7]:
In [8]:
df1.plot(kind='box')
Out[8]:
In [10]:
df1['steering'].plot(kind='kde')
Out[10]:
In [9]:
# normalise data
df1_norm = normalize(df1[df1_summary_cols][:])
# put into dataframe for easy charting
df1_norm = pd.DataFrame(df1_norm, columns=df1_summary_cols)
# drop columns with zero mean and zero std.dev
df1_norm = df1_norm.loc[:, (df1_norm.mean() != 0.) & (df1_norm.std() != 0.)]
# plot kde
df1_norm.plot(kind='kde',
subplots=True,
figsize=(10, 5))
plt.tight_layout()
In [10]:
df2 = pd.read_csv('../data/raw/sim_track_1_direction_default/driving_log.csv',
header=None,
names=columns)
In [11]:
df2_count = len(df2)
df2_count
Out[11]:
In [12]:
df2.info()
In [13]:
df2.head()
Out[13]:
In [14]:
print(df2['center'][0])
In [15]:
!pwd
In [16]:
df2_summary = df2.describe().transpose()
df2_summary_cols = df2_summary.index
print(df2_summary_cols)
df2_summary
Out[16]:
In [17]:
df2.plot(kind='box')
Out[17]:
In [18]:
# normalise data
df2_norm = normalize(df2[df2_summary_cols][:])
# put into dataframe for easy charting
df2_norm = pd.DataFrame(df2_norm, columns=df2_summary_cols)
# drop columns with zero mean and zero std.dev
df2_norm = df2_norm.loc[:, (df2_norm.mean() != 0.) & (df2_norm.std() != 0.)]
# plot kde
df2_norm.plot(kind='kde',
subplots=True,
figsize=(10, 5))
plt.tight_layout()
In [19]:
df3 = pd.read_csv('../data/raw/sim_track_1_direction_reverse/driving_log.csv',
header=None,
names=columns)
In [20]:
df3_count = len(df3)
df3_count
Out[20]:
In [21]:
df3.info()
In [22]:
df3.head()
Out[22]:
In [23]:
df3_summary = df3.describe().transpose()
df3_summary_cols = df3_summary.index
print(df3_summary_cols)
df3_summary
Out[23]:
In [24]:
df3.plot(kind='box')
Out[24]:
In [25]:
# normalise data
df3_norm = normalize(df3[df3_summary_cols][:])
# put into dataframe for easy charting
df3_norm = pd.DataFrame(df3_norm, columns=df3_summary_cols)
# drop columns with zero mean and zero std.dev
df3_norm = df3_norm.loc[:, (df3_norm.mean() != 0.) & (df3_norm.std() != 0.)]
# plot kde
df3_norm.plot(kind='kde',
subplots=True,
figsize=(10, 5))
plt.tight_layout()
In [26]:
df4 = pd.read_csv('../data/raw/sim_track_2_direction_default/driving_log.csv',
header=None,
names=columns)
In [27]:
df4_count = len(df3)
df4_count
Out[27]:
In [28]:
df4.info()
In [29]:
df4.head()
Out[29]:
In [30]:
df4_summary = df4.describe().transpose()
df4_summary_cols = df4_summary.index
print(df4_summary_cols)
df4_summary
Out[30]:
In [31]:
df4.plot(kind='box')
Out[31]:
In [32]:
# normalise data
df4_norm = normalize(df4[df4_summary_cols][:])
# put into dataframe for easy charting
df4_norm = pd.DataFrame(df4_norm, columns=df4_summary_cols)
# drop columns with zero mean and zero std.dev
df4_norm = df3_norm.loc[:, (df4_norm.mean() != 0.) & (df4_norm.std() != 0.)]
# plot kde
df4_norm.plot(kind='kde',
subplots=True,
figsize=(10, 5))
plt.tight_layout()
In [33]:
df5 = pd.read_csv('../data/raw/sim_track_2_direction_reverse/driving_log.csv',
header=None,
names=columns)
In [34]:
df5_count = len(df5)
df5_count
Out[34]:
In [35]:
df5.info()
In [36]:
df5.head()
Out[36]:
In [37]:
df5_summary = df5.describe().transpose()
df5_summary_cols = df5_summary.index
print(df5_summary_cols)
df5_summary
Out[37]:
In [38]:
df5.plot(kind='box')
Out[38]:
In [39]:
# normalise data
df5_norm = normalize(df5[df5_summary_cols][:])
# put into dataframe for easy charting
df5_norm = pd.DataFrame(df5_norm, columns=df5_summary_cols)
# drop columns with zero mean and zero std.dev
df5_norm = df5_norm.loc[:, (df5_norm.mean() != 0.) & (df5_norm.std() != 0.)]
# plot kde
df5_norm.plot(kind='kde',
subplots=True,
figsize=(10, 5))
plt.tight_layout()
In [40]:
track_1_count = df2_count + df3_count
track_2_count = df4_count + df5_count
total_count = df1_count + df2_count + df3_count + df4_count + df5_count
print('source\t\t\t\t\t n\t total\t %')
print('==================================== ===== ======\t ======\n')
print('sample: total \t\t\t\t{:>6,}\t\t {:>5.3}%'.format(df1_count,
(df1_count / total_count) * 100))
print('----------------------------------------------------------------\n')
print('sim track 1: direction=default \t\t{:>6,}'.format(df2_count))
print('sim track 1: direction=reverse \t\t{:>6,}'.format(df3_count))
print('sim track 1: total \t\t\t\t {:>6,}\t {:>5.3}%'.format(track_1_count,
(track_1_count / total_count) * 100))
print('----------------------------------------------------------------\n')
print('sim track 2: direction=default \t\t{:>6,}'.format(df4_count))
print('sim track 2: direction=reverse \t\t{:>6,}'.format(df5_count))
print('sim track 2: total \t\t\t\t {:>6,}\t {:>5.3}%'.format(track_2_count,
(track_2_count / total_count) * 100))
print('----------------------------------------------------------------\n')
print('totals (x3 for centre + left + right)\t\t {:>6,}\t 100.0%'.format(total_count))
print('================================================================')
In [51]:
img_path = '/src/repos/udacity_carnd/project_3/data/raw/sample/IMG/center_2016_12_01_13_30_48_287.jpg'
img = cv2.imread(img_path, cv2.COLOR_BGR2RGB)
plt.imshow(img)
Out[51]:
In [52]:
type(img)
Out[52]:
In [53]:
img = img[60:140, :, :]
img.shape
Out[53]:
In [54]:
plt.imshow(img)
Out[54]: