In this notebook, a template is provided for you to implement your functionality in stages, which is required to successfully complete this project. If additional code is required that cannot be included in the notebook, be sure that the Python code is successfully imported and included in your submission if necessary.
Note: Once you have completed all of the code implementations, you need to finalize your work by exporting the iPython Notebook as an HTML document. Before exporting the notebook to html, all of the code cells need to have been run so that reviewers can see the final implementation and output. You can then export the notebook by using the menu above and navigating to \n", "File -> Download as -> HTML (.html). Include the finished document along with this notebook as your submission.
In addition to implementing code, there is a writeup to complete. The writeup should be completed in a separate file, which can be either a markdown file or a pdf document. There is a write up template that can be used to guide the writing process. Completing the code template and writeup template will cover all of the rubric points for this project.
The rubric contains "Stand Out Suggestions" for enhancing the project beyond the minimum requirements. The stand out suggestions are optional. If you decide to pursue the "stand out suggestions", you can include the code in this Ipython notebook and also discuss the results in the writeup file.
Note: Code and Markdown cells can be executed using the Shift + Enter keyboard shortcut. In addition, Markdown cells can be edited by typically double-clicking the cell to enter edit mode.
In [1]:
# Load pickled data
import pickle
# TODO: Fill this in based on where you saved the training and testing data
training_file = './traffic-signs-data/train.p'
validation_file = './traffic-signs-data/valid.p'
testing_file = './traffic-signs-data/test.p'
with open(training_file, mode='rb') as f:
train = pickle.load(f)
with open(validation_file, mode='rb') as f:
valid = pickle.load(f)
with open(testing_file, mode='rb') as f:
test = pickle.load(f)
X_train, y_train = train['features'], train['labels']
X_valid, y_valid = valid['features'], valid['labels']
X_test, y_test = test['features'], test['labels']
assert(len(X_train) == len(y_train))
assert(len(X_valid) == len(y_valid))
assert(len(X_test) == len(y_test))
print("Loading done!")
The pickled data is a dictionary with 4 key/value pairs:
'features' is a 4D array containing raw pixel data of the traffic sign images, (num examples, width, height, channels).'labels' is a 1D array containing the label/class id of the traffic sign. The file signnames.csv contains id -> name mappings for each id.'sizes' is a list containing tuples, (width, height) representing the original width and height the image.'coords' is a list containing tuples, (x1, y1, x2, y2) representing coordinates of a bounding box around the sign in the image. THESE COORDINATES ASSUME THE ORIGINAL IMAGE. THE PICKLED DATA CONTAINS RESIZED VERSIONS (32 by 32) OF THESE IMAGESComplete the basic data summary below. Use python, numpy and/or pandas methods to calculate the data summary rather than hard coding the results. For example, the pandas shape method might be useful for calculating some of the summary results.
In [2]:
### Replace each question mark with the appropriate value.
### Use python, pandas or numpy methods rather than hard coding the results
import numpy as np
# Number of training examples
n_train = len(X_train)
# Number of testing examples.
n_test = len(X_test)
# Number of validation examples
n_valid = len(X_valid)
# TODO: What's the shape of an traffic sign image?
image_shape = X_train[0].shape
# TODO: How many unique classes/labels there are in the dataset.
n_classes = np.unique(y_train).size
print("Number of training examples =", n_train)
print("Number of validation examples =", n_valid)
print("Number of testing examples =", n_test)
print("Image data shape =", image_shape)
print("Number of classes =", n_classes)
Visualize the German Traffic Signs Dataset using the pickled file(s). This is open ended, suggestions include: plotting traffic sign images, plotting the count of each sign, etc.
The Matplotlib examples and gallery pages are a great resource for doing visualizations in Python.
NOTE: It's recommended you start with something simple first. If you wish to do more, come back to it after you've completed the rest of the sections.
In [3]:
import matplotlib.pyplot as plt
import random
import numpy as np
import csv
import pandas as pd
# Visualizations will be shown in the notebook.
%matplotlib inline
def show_sample(features, labels, histogram = 1, sample_num = 1, sample_index = -1, color_map ='brg'):
if histogram == 1 :
col_num = 2
#Create training sample + histogram plot
f, axarr = plt.subplots(sample_num+1, col_num, figsize=(col_num*4,(sample_num+1)*3))
else:
if sample_num <= 4:
col_num = sample_num
else:
col_num = 4
if sample_num%col_num == 0:
row_num = int(sample_num/col_num)
else:
row_num = int(sample_num/col_num)+1
if sample_num == 1:
#Create training sample plot
f, ax = plt.subplots(row_num, col_num)
else:
#Create training sample plot
f, axarr = plt.subplots(row_num, col_num, figsize=(col_num*4,(row_num+1)*2))
signnames = pd.read_csv('signnames.csv')
index = sample_index - 1
for i in range(0, sample_num, 1):
if sample_index < -1:
index = random.randint(0, len(features))
else:
index = index + 1
if histogram == 1 :
image = features[index].squeeze()
axarr[i,0].set_title('%s' % signnames.iloc[labels[index], 1])
axarr[i,0].imshow(image,color_map)
hist,bins = np.histogram(image.flatten(),256, normed =1 )
cdf = hist.cumsum()
cdf_normalized = cdf * hist.max()/ cdf.max()
axarr[i,1].plot(cdf_normalized, color = 'b')
axarr[i,1].hist(image.flatten(),256, normed =1, color = 'r')
axarr[i,1].legend(('cdf','histogram'), loc = 'upper left')
axarr[i,0].axis('off')
axarr[sample_num,0].axis('off')
axarr[sample_num,1].axis('off')
else:
image = features[index].squeeze()
if row_num > 1:
axarr[int(i/col_num),i%col_num].set_title('%s' % signnames.iloc[labels[index], 1])
axarr[int(i/col_num),i%col_num].imshow(image,color_map)
axarr[int(i/col_num),i%col_num].axis('off')
axarr[int(i/col_num),i%col_num].axis('off')
axarr[int(i/col_num),i%col_num].axis('off')
elif sample_num == 1:
ax.set_title('%s' % signnames.iloc[labels[index], 1])
ax.imshow(image,color_map)
ax.axis('off')
ax.axis('off')
ax.axis('off')
else:
axarr[i%col_num].set_title('%s' % signnames.iloc[labels[index], 1])
axarr[i%col_num].imshow(image,color_map)
axarr[i%col_num].axis('off')
axarr[i%col_num].axis('off')
axarr[i%col_num].axis('off')
# Tweak spacing to prevent clipping of title labels
f.tight_layout()
plt.show()
def show_training_dataset_histogram(labels_train,labels_valid,labels_test):
fig, ax = plt.subplots(figsize=(15,5))
temp = [labels_train,labels_valid,labels_test]
n_classes = np.unique(y_train).size
# the histogram of the training data
n, bins, patches = ax.hist(temp, n_classes, label=["Train","Valid","Test"])
ax.set_xlabel('Classes')
ax.set_ylabel('Number of occurence')
ax.set_title(r'Histogram of the data sets')
ax.legend(bbox_to_anchor=(1.01, 1), loc="upper left")
plt.show()
show_training_dataset_histogram(y_train,y_valid,y_test)
show_sample(X_train, y_train, sample_num = 6)
Design and implement a deep learning model that learns to recognize traffic signs. Train and test your model on the German Traffic Sign Dataset.
The LeNet-5 implementation shown in the classroom at the end of the CNN lesson is a solid starting point. You'll have to change the number of classes and possibly the preprocessing, but aside from that it's plug and play!
With the LeNet-5 solution from the lecture, you should expect a validation set accuracy of about 0.89. To meet specifications, the validation set accuracy will need to be at least 0.93. It is possible to get an even higher accuracy, but 0.93 is the minimum for a successful project submission.
There are various aspects to consider when thinking about this problem:
Here is an example of a published baseline model on this problem. It's not required to be familiar with the approach used in the paper but, it's good practice to try to read papers like these.
In [4]:
import cv2
from tqdm import tqdm
from sklearn.utils import shuffle
def random_transform_image(dataset, index):
# Hyperparameters
# Values inspired from Pierre Sermanet and Yann LeCun Paper : Traffic Sign Recognition with Multi-Scale Convolutional Networks
Scale_change_max = 0.1
Translation_max = 2 #pixels
Rotation_max = 15 #degrees
Brightness_max = 0.1
# Generate random transformation values
trans_x = np.random.uniform(-Translation_max,Translation_max)
trans_y = np.random.uniform(-Translation_max,Translation_max)
angle = np.random.uniform(-Rotation_max,Rotation_max)
scale = np.random.uniform(1-Scale_change_max,1+Scale_change_max)
bright = np.random.uniform(-Brightness_max,Brightness_max)
#Brightness
#create white image
white_img = 255*np.ones((32,32,3), np.uint8)
black_img = np.zeros((32,32,3), np.uint8)
if bright >= 0:
img = cv2.addWeighted(dataset[index].squeeze(),1-bright,white_img,bright,0)
else:
img = cv2.addWeighted(dataset[index].squeeze(),bright+1,black_img,bright*-1,0)
# Scale
img = cv2.resize(img,None,fx=scale, fy=scale, interpolation = cv2.INTER_CUBIC)
# Get image shape afeter scaling
rows,cols,chan = img.shape
# Pad with zeroes before rotation if image shape is less than 32*32*3
if rows < 32:
offset = int((32-img.shape[0])/2)
# If shape is an even number
if img.shape[0] %2 == 0:
img = cv2.copyMakeBorder(img,offset,offset,offset,offset,cv2.BORDER_CONSTANT,value=[0,0,0])
else:
img = cv2.copyMakeBorder(img,offset,offset+1,offset+1,offset,cv2.BORDER_CONSTANT,value=[0,0,0])
# Update image shape after padding
rows,cols,chan = img.shape
# Rotate
M = cv2.getRotationMatrix2D((cols/2,rows/2),angle,1)
img = cv2.warpAffine(img,M,(cols,rows))
# Translation
M = np.float32([[1,0,trans_x],[0,1,trans_y]])
img = cv2.warpAffine(img,M,(cols,rows))
# Crop centered if image shape is greater than 32*32*3
if rows > 32:
offset = int((img.shape[0]-32)/2)
img = img[offset: 32 + offset, offset: 32 + offset]
return img
# Parameters
# Max example number per class
num_example_per_class = np.bincount(y_train)
min_example_num = max(num_example_per_class)
for i in range(len(num_example_per_class)):
# Update number of examples by class
num_example_per_class = np.bincount(y_train)
# If the class lacks examples...
if num_example_per_class[i] < min_example_num:
# Locate where pictures of this class are located in the training set..
pictures = np.array(np.where(y_train == i)).T
# Compute the number of pictures to be generated
num_example_to_generate = min_example_num - num_example_per_class[i]
# Compute the number of iteration necessary on the real data
num_iter = int( num_example_to_generate/len(pictures) ) + 1
# Compute the pool of real data necessary to fill the classes
if num_iter == 1 :
num_pictures = num_example_to_generate
else:
num_pictures = len(pictures)
# # Limit the number of iteration to 10
# num_iter = min(num_iter, 10)
# Create empty list
more_X = []
more_y = []
for k in range(num_iter):
# if we are in the last iteration, num_pictures is adjusted to fit the min_example_num
if (k == num_iter - 1) and (num_iter > 1):
num_pictures = min_example_num - num_iter * len(pictures)
# For each pictures of this class, generate 1 more synthetic image
pbar = tqdm(range(num_pictures), desc='Iter {:>2}/{}'.format(i+1, len(num_example_per_class)), unit='examples')
for j in pbar:
# Append the transformed picture
more_X.append(random_transform_image(X_train,pictures[j]))
# Append the class number
more_y.append(i)
# Append the synthetic images to the training set
X_train = np.append(X_train, np.array(more_X), axis=0)
y_train = np.append(y_train, np.array(more_y), axis=0)
print("New training feature shape",X_train.shape)
print("New training label shape",y_train.shape)
print("Data augmentation done!")
In [5]:
# Visualization
show_training_dataset_histogram(y_train,y_valid,y_test)
show_sample(X_train, y_train, histogram = 0, sample_num = 8, sample_index = 35000)
In [6]:
import cv2
from numpy import newaxis
def equalize_Y_histogram(features):
images = []
for image in features:
# Convert RGB to YUV
temp = cv2.cvtColor(image, cv2.COLOR_BGR2YUV);
# Equalize Y histogram in order to get better contrast accross the dataset
temp[:,:,0] = cv2.equalizeHist(temp[:,:,0])
# Convert back YUV to RGB
temp = cv2.cvtColor(temp, cv2.COLOR_YUV2BGR)
images.append(temp)
return np.array(images)
def CLAHE_contrast_normalization(features):
images = []
for image in features:
# create a CLAHE object
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(4,4))
temp = clahe.apply(image)
images.append(temp)
return np.array(images)
def convert_to_grayscale(features):
gray_images = []
for image in features:
# Convert RGB to grayscale
temp = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
gray_images.append(temp)
return np.array(gray_images)
def normalize_grayscale(image_data):
"""
Normalize the image data with Min-Max scaling to a range of [0.1, 0.9]
:param image_data: The image data to be normalized
:return: Normalized image data
"""
a = 0.1
b = 0.9
image_data_norm = a + ((image_data - np.amin(image_data))*(b-a))/(np.amax(image_data) - np.amin(image_data))
return image_data_norm
In [7]:
index = 255
X_temp1 = convert_to_grayscale(X_train)
X_temp2 = CLAHE_contrast_normalization(X_temp1)
X_temp3 = normalize_grayscale(X_temp2)
show_sample(X_train, y_train, histogram = 1, sample_num = 1, sample_index = index)
show_sample(X_temp1, y_train, histogram = 1, sample_num = 1, sample_index = index, color_map ='gray')
show_sample(X_temp2, y_train, histogram = 1, sample_num = 1, sample_index = index, color_map ='gray')
print(X_temp2[index])
print(X_temp3[index])
In [8]:
#Preprocessing pipeline
print('Preprocessing training features...')
X_train = convert_to_grayscale(X_train)
X_train = CLAHE_contrast_normalization(X_train)
X_train = normalize_grayscale(X_train)
X_train = X_train[..., newaxis]
print("Processed shape =", X_train.shape)
print('Preprocessing validation features...')
X_valid = convert_to_grayscale(X_valid)
X_valid = CLAHE_contrast_normalization(X_valid)
X_valid = normalize_grayscale(X_valid)
X_valid = X_valid[..., newaxis]
print("Processed shape =", X_valid.shape)
print('Preprocessing test features...')
X_test = convert_to_grayscale(X_test)
X_test = CLAHE_contrast_normalization(X_test)
X_test = normalize_grayscale(X_test)
X_test = X_test[..., newaxis]
print("Processed shape =", X_test.shape)
# Shuffle the training dataset
X_train, y_train = shuffle(X_train, y_train)
print("Pre-processing done!")
| Layer | Description | Input | Output |
|---|---|---|---|
| Input | 32x32x1 Grayscale image | Image | Convolution 1 |
| Convolution 1 | 1x1 stride, valid padding, outputs 28x28x100 | Input | RELU |
| RELU 1 | Convolution 1 | Max Pooling 1 | |
| Max pooling 1 | 2x2 stride, outputs 14x14x100 | RELU 1 | Convolution 2, Max Pooling 3 |
| Convolution 2 | 1x1 stride, valid padding, outputs 10x10x200 | Max pooling 1 | RELU 2 |
| RELU 2 | Convolution 2 | Max pooling 2 | |
| Max pooling 2 | 2x2 stride, outputs 5x5x200 | RELU 2 | Flatten 2 |
| Max pooling 3 | 2x2 stride, outputs 7x7x100 | Max pooling 1 | Flatten 1 |
| Flatten 1 | Input = 7x7x100, Output = 4900 | Max pooling 3 | Concatenate 1 |
| Flatten 2 | Input = 5x5x200, Output = 5000 | Max pooling 2 | Concatenate 1 |
| Concatenate 1 | Input1 = 4900, Input1 = 5000, Output = 9900 | Max pooling 2 and 3 | Fully connected |
| Fully connected | Fully Connected. Input = 9900, Output = 100 | Concatenate 1 | Dropout |
| Dropout | Keep prob = 0.75 | Fully connected | Softmax |
| Softmax | Fully Connected. Input = 100, Output = 43 | Dropout | Probabilities |
In [9]:
import tensorflow as tf
from tensorflow.contrib.layers import flatten
def model(x, keep_prob):
# Arguments used for tf.truncated_normal, randomly defines variables for the weights and biases for each layer
mu = 0
sigma = 0.1
# Network Parameters
n_classes = 43 # MNIST total classes (0-9 digits)
filter_size = 5
# Store layers weight & bias
weights = {
'wc1' : tf.Variable(tf.truncated_normal([filter_size, filter_size, 1, 100], mean = mu, stddev = sigma)),
'wc2' : tf.Variable(tf.truncated_normal([filter_size, filter_size, 100, 200], mean = mu, stddev = sigma)),
'wfc1': tf.Variable(tf.truncated_normal([9900, 100], mean = mu, stddev = sigma)),
'out' : tf.Variable(tf.truncated_normal([100, n_classes], mean = mu, stddev = sigma))}
biases = {
'bc1' : tf.Variable(tf.zeros([100])),
'bc2' : tf.Variable(tf.zeros([200])),
'bfc1': tf.Variable(tf.zeros([100])),
'out' : tf.Variable(tf.zeros([n_classes]))}
def conv2d(x, W, b, strides=1., padding='SAME'):
x = tf.nn.conv2d(x, W, strides=[1, strides, strides, 1], padding=padding)
x = tf.nn.bias_add(x, b)
return tf.nn.relu(x)
def maxpool2d(x, k=2, padding='SAME'):
return tf.nn.max_pool(x, ksize=[1, k, k, 1], strides=[1, k, k, 1], padding=padding)
# Layer 1: Convolution 1 - 32*32*1 to 28*28*100
conv1 = conv2d(x, weights['wc1'], biases['bc1'], padding='VALID')
# Max Pool - 28*28*100 to 14*14*100
conv1 = maxpool2d(conv1, k=2)
# Layer 2: Convolution 2 - 14*14*100 to 10*10*200
conv2 = conv2d(conv1, weights['wc2'], biases['bc2'], padding='VALID')
# Max Pool - 10*10*200 to 5*5*200
conv2 = maxpool2d(conv2, k=2)
#Fork second max pool - 14*14*100 to 7*7*100
conv1 = maxpool2d(conv1, k=2)
#Flatten conv1. Input = 7*7*100, Output = 4900
conv1 = tf.contrib.layers.flatten(conv1)
# Flatten conv2. Input = 5x5x200. Output = 5000.
conv2 = tf.contrib.layers.flatten(conv2)
# Concatenate
flat = tf.concat(1,[conv1,conv2])
# Layer 3 : Fully Connected. Input = 9900. Output = 100.
fc1 = tf.add(tf.matmul(flat, weights['wfc1']), biases['bfc1'])
fc1 = tf.nn.relu(fc1)
fc1 = tf.nn.dropout(fc1, keep_prob)
# Layer 4: Fully Connected. Input = 100. Output = 43.
logits = tf.add(tf.matmul(fc1, weights['out']), biases['out'])
return logits
A validation set can be used to assess how well the model is performing. A low accuracy on the training and validation sets imply underfitting. A high accuracy on the training set but low accuracy on the validation set implies overfitting.
In [10]:
### Train your model here.
### Calculate and report the accuracy on the training and validation set.
### Once a final model architecture is selected,
### the accuracy on the test set should be calculated and reported as well.
### Feel free to use as many code cells as needed.
#Hyperparameters
EPOCHS = 100 #Max EPOCH number, if ever early stopping doesn't kick in
BATCH_SIZE = 256 #Max batch size
rate = 0.001 #Base learning rate
keep_probability = 0.75 #Keep probability for dropout..
max_iter_wo_improvmnt = 3000 #For early stopping
In [11]:
#Declare placeholder tensors
x = tf.placeholder(tf.float32, (None, 32, 32, 1))
y = tf.placeholder(tf.int32, (None))
keep_prob = tf.placeholder(tf.float32)
one_hot_y = tf.one_hot(y, 43)
In [12]:
logits = model(x, keep_prob)
probabilities = tf.nn.softmax(logits)
cross_entropy = tf.nn.softmax_cross_entropy_with_logits(logits, one_hot_y)
loss_operation = tf.reduce_mean(cross_entropy)
optimizer = tf.train.AdamOptimizer(learning_rate = rate)
training_operation = optimizer.minimize(loss_operation)
In [13]:
correct_prediction = tf.equal(tf.argmax(logits, 1), tf.argmax(one_hot_y, 1))
accuracy_operation = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
saver = tf.train.Saver()
def evaluate(X_data, y_data):
num_examples = len(X_data)
total_accuracy = 0
sess = tf.get_default_session()
for offset in range(0, num_examples, BATCH_SIZE):
batch_x, batch_y = X_data[offset:offset+BATCH_SIZE], y_data[offset:offset+BATCH_SIZE]
accuracy = sess.run(accuracy_operation, feed_dict={x: batch_x, y: batch_y, keep_prob: 1.0})
total_accuracy += (accuracy * len(batch_x))
return total_accuracy / num_examples
In [16]:
from sklearn.utils import shuffle
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
num_examples = len(X_train)
# Max iteration number without improvement
max_interation_num_wo_improv = 1000
print("Training...")
iteration = 0
best_valid_accuracy = 0
best_accuracy_iter = 0
stop = 0
print()
for i in range(EPOCHS):
X_train, y_train = shuffle(X_train, y_train)
for offset in range(0, num_examples, BATCH_SIZE):
iteration = iteration + 1
end = offset + BATCH_SIZE
batch_x, batch_y = X_train[offset:end], y_train[offset:end]
sess.run(training_operation, feed_dict={x: batch_x, y: batch_y, keep_prob: keep_probability})
# After 10 Epochs, for every 200 iterations validation accuracy is checked
if (iteration % 200 == 0 and i > 10):
validation_accuracy = evaluate(X_valid, y_valid)
if validation_accuracy > best_valid_accuracy:
best_valid_accuracy = validation_accuracy
best_accuracy_iter = iteration
saver = tf.train.Saver()
saver.save(sess, './best_model')
print("Improvement found, model saved!")
stop = 0
# Stopping criteria : if not improvement since 1000 iterations stop training
if (iteration - best_accuracy_iter) > max_iter_wo_improvmnt:
print("Stopping criteria met..")
stop = 1
validation_accuracy = evaluate(X_valid, y_valid)
print("EPOCH {} ...".format(i+1))
print("Validation Accuracy = {:.3f}".format(validation_accuracy))
print()
if stop == 1:
break
# saver.save(sess, './lenet')
# print("Model saved")
In [17]:
### Load the images and plot them here.
### Feel free to use as many code cells as needed.
with tf.Session() as sess:
saver.restore(sess, tf.train.latest_checkpoint('.'))
print("Evaluating..")
train_accuracy = evaluate(X_train, y_train)
print("Train Accuracy = {:.3f}".format(train_accuracy))
valid_accuracy = evaluate(X_valid, y_valid)
print("Valid Accuracy = {:.3f}".format(valid_accuracy))
test_accuracy = evaluate(X_test, y_test)
print("Test Accuracy = {:.3f}".format(test_accuracy))
To give yourself more insight into how your model is working, download at least five pictures of German traffic signs from the web and use your model to predict the traffic sign type.
You may find signnames.csv useful as it contains mappings from the class id (integer) to the actual sign name.
In [14]:
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
import os
test_images = os.listdir('traffic-signs-data/web_found_signs/')
X_web = []
for file in test_images:
image = mpimg.imread('traffic-signs-data/web_found_signs/' + file)
plt.imshow(image)
plt.show()
print("Loaded ", file)
X_web.append(image)
X_web = np.array(X_web)
# Preprocess images
print('Preprocessing features...')
X_web = equalize_Y_histogram(X_web)
X_web = convert_to_grayscale(X_web)
X_web = normalize_grayscale(X_web)
X_web = X_web[..., newaxis]
print("Processed shape =", X_web.shape)
In [15]:
### Run the predictions here and use the model to output the prediction for each image.
### Make sure to pre-process the images with the same pre-processing pipeline used earlier.
### Feel free to use as many code cells as needed.
import tensorflow as tf
# hardcoded..
y_web = [9,22,2,18,1,17,4,10,38,4,4,23]
#We have to set the keep probability to 1.0 in the model..
with tf.Session() as sess:
saver.restore(sess, tf.train.latest_checkpoint('.'))
logits_web = sess.run(tf.argmax(logits,1), feed_dict={x: X_web, keep_prob: 1.0})
print("Prediction =", logits_web)
# show_sample(X_web, logits_web, histogram = 0, sample_num = len(test_images), sample_index = 0, color_map = 'gray')
#Number of column to show
sample_num = len(test_images)
col_num = 4
if sample_num%col_num == 0:
row_num = int(sample_num/col_num)
else:
row_num = int(sample_num/col_num)+1
#Create training sample plot
f, axarr = plt.subplots(row_num, col_num, figsize=(col_num*4,(row_num+1)*2))
signnames = pd.read_csv('signnames.csv')
for i in range(0, sample_num, 1):
image = X_web[i].squeeze()
if logits_web[i] != y_web[i]:
color_str = 'red'
else:
color_str = 'green'
title_str = 'Predicted : %s \n Real: %s' % (signnames.iloc[logits_web[i], 1],signnames.iloc[y_web[i], 1])
axarr[int(i/col_num),i%col_num].set_title(title_str, color = color_str)
axarr[int(i/col_num),i%col_num].imshow(image,'gray')
axarr[int(i/col_num),i%col_num].axis('off')
axarr[int(i/col_num),i%col_num].axis('off')
axarr[int(i/col_num),i%col_num].axis('off')
f.tight_layout()
plt.show()
In [16]:
### Calculate the accuracy for these 5 new images.
with tf.Session() as sess:
saver.restore(sess, tf.train.latest_checkpoint('.'))
test_accuracy = evaluate(X_web, y_web)
print("Web images Accuracy = {:.3f}".format(test_accuracy))
For each of the new images, print out the model's softmax probabilities to show the certainty of the model's predictions (limit the output to the top 5 probabilities for each image). tf.nn.top_k could prove helpful here.
The example below demonstrates how tf.nn.top_k can be used to find the top k predictions for each image.
tf.nn.top_k will return the values and indices (class ids) of the top k predictions. So if k=3, for each sign, it'll return the 3 largest probabilities (out of a possible 43) and the correspoding class ids.
Take this numpy array as an example. The values in the array represent predictions. The array contains softmax probabilities for five candidate images with six possible classes. tk.nn.top_k is used to choose the three classes with the highest probability:
# (5, 6) array
a = np.array([[ 0.24879643, 0.07032244, 0.12641572, 0.34763842, 0.07893497,
0.12789202],
[ 0.28086119, 0.27569815, 0.08594638, 0.0178669 , 0.18063401,
0.15899337],
[ 0.26076848, 0.23664738, 0.08020603, 0.07001922, 0.1134371 ,
0.23892179],
[ 0.11943333, 0.29198961, 0.02605103, 0.26234032, 0.1351348 ,
0.16505091],
[ 0.09561176, 0.34396535, 0.0643941 , 0.16240774, 0.24206137,
0.09155967]])Running it through sess.run(tf.nn.top_k(tf.constant(a), k=3)) produces:
TopKV2(values=array([[ 0.34763842, 0.24879643, 0.12789202],
[ 0.28086119, 0.27569815, 0.18063401],
[ 0.26076848, 0.23892179, 0.23664738],
[ 0.29198961, 0.26234032, 0.16505091],
[ 0.34396535, 0.24206137, 0.16240774]]), indices=array([[3, 0, 5],
[0, 1, 4],
[0, 5, 1],
[1, 3, 5],
[1, 4, 3]], dtype=int32))Looking just at the first row we get [ 0.34763842, 0.24879643, 0.12789202], you can confirm these are the 3 largest probabilities in a. You'll also notice [3, 0, 5] are the corresponding indices.
In [17]:
### Print out the top five softmax probabilities for the predictions on the German traffic sign images found on the web.
### Feel free to use as many code cells as needed.
import matplotlib.gridspec as gridspec
with tf.Session() as sess:
saver.restore(sess, tf.train.latest_checkpoint('.'))
softmax_prob = sess.run(tf.nn.top_k(probabilities,k = 5), feed_dict={x: X_web, keep_prob: 1.0})
signnames = pd.read_csv('signnames.csv')
for i in range(len(test_images)):
plt.figure(figsize = (6,2))
gs = gridspec.GridSpec(1, 2,width_ratios=[2,3])
plt.subplot(gs[0])
plt.imshow(X_web[i].squeeze(),cmap="gray")
plt.axis('off')
plt.subplot(gs[1])
plt.barh(6-np.arange(5),softmax_prob[0][i], align='center')
if logits_web[i] != y_web[i]:
color_str = 'red'
else:
color_str = 'green'
for i_label in range(5):
temp_string = "%.1f %% : %s" % (softmax_prob[0][i][i_label]*100, str(signnames.iloc[softmax_prob[1][i][i_label], 1]))
plt.text(softmax_prob[0][i][0]*1.1,6-i_label-.15, temp_string, color = color_str)
plt.show()
This Section is not required to complete but acts as an additional excersise for understaning the output of a neural network's weights. While neural networks can be a great learning device they are often referred to as a black box. We can understand what the weights of a neural network look like better by plotting their feature maps. After successfully training your neural network you can see what it's feature maps look like by plotting the output of the network's weight layers in response to a test stimuli image. From these plotted feature maps, it's possible to see what characteristics of an image the network finds interesting. For a sign, maybe the inner network feature maps react with high activation to the sign's boundary outline or to the contrast in the sign's painted symbol.
Provided for you below is the function code that allows you to get the visualization output of any tensorflow weight layer you want. The inputs to the function should be a stimuli image, one used during training or a new one you provided, and then the tensorflow variable name that represents the layer's state during the training process, for instance if you wanted to see what the LeNet lab's feature maps looked like for it's second convolutional layer you could enter conv2 as the tf_activation variable.
For an example of what feature map outputs look like, check out NVIDIA's results in their paper End-to-End Deep Learning for Self-Driving Cars in the section Visualization of internal CNN State. NVIDIA was able to show that their network's inner weights had high activations to road boundary lines by comparing feature maps from an image with a clear path to one without. Try experimenting with a similar test to show that your trained network's weights are looking for interesting features, whether it's looking at differences in feature maps from images with or without a sign, or even what feature maps look like in a trained network vs a completely untrained one on the same sign image.
Your output should look something like this (above)
In [22]:
### Visualize your network's feature maps here.
### Feel free to use as many code cells as needed.
# image_input: the test image being fed into the network to produce the feature maps
# tf_activation: should be a tf variable name used during your training procedure that represents the calculated state of a specific weight layer
# activation_min/max: can be used to view the activation contrast in more detail, by default matplot sets min and max to the actual min and max values of the output
# plt_num: used to plot out multiple different weight feature map sets on the same block, just extend the plt number for each new feature map entry
def outputFeatureMap(image_input, tf_activation, activation_min=-1, activation_max=-1 ,plt_num=1):
# Here make sure to preprocess your image_input in a way your network expects
# with size, normalization, ect if needed
# image_input =
# Note: x should be the same name as your network's tensorflow data placeholder variable
# If you get an error tf_activation is not defined it maybe having trouble accessing the variable from inside a function
activation = tf_activation.eval(session=sess,feed_dict={x : image_input})
featuremaps = activation.shape[3]
plt.figure(plt_num, figsize=(15,15))
for featuremap in range(featuremaps):
plt.subplot(6,8, featuremap+1) # sets the number of feature maps to show on each row and column
plt.title('FeatureMap ' + str(featuremap)) # displays the feature map number
if activation_min != -1 & activation_max != -1:
plt.imshow(activation[0,:,:, featuremap], interpolation="nearest", vmin =activation_min, vmax=activation_max, cmap="gray")
elif activation_max != -1:
plt.imshow(activation[0,:,:, featuremap], interpolation="nearest", vmax=activation_max, cmap="gray")
elif activation_min !=-1:
plt.imshow(activation[0,:,:, featuremap], interpolation="nearest", vmin=activation_min, cmap="gray")
else:
plt.imshow(activation[0,:,:, featuremap], interpolation="nearest", cmap="gray")
Answer:
Note: Once you have completed all of the code implementations and successfully answered each question above, you may finalize your work by exporting the iPython Notebook as an HTML document. You can do this by using the menu above and navigating to \n", "File -> Download as -> HTML (.html). Include the finished document along with this notebook as your submission.
Once you have completed the code implementation, document your results in a project writeup using this template as a guide. The writeup can be in a markdown or pdf file.