Self-Driving Car Engineer Nanodegree¶
Deep Learning¶
Project: Build a Traffic Sign Recognition Classifier¶
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.
Step 0: Load The Data¶
# Load pickled data
import pickle
import numpy as np
import tensorflow as tf
import pandas as pd
import matplotlib.pyplot as plt
import cv2
import os
from sklearn.utils import shuffle
from sklearn.metrics import confusion_matrix
# Visualizations will be shown in the notebook.
%matplotlib inline
training_file = "data/train.p"
validation_file= "data/valid.p"
testing_file = "data/test.p"
signnames_file = "signnames.csv"
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)
with open(signnames_file) as f:
f.readline() # Strip the header
tuples = [line.strip().split(',') for line in f]
sign_names = {int(t[0]): t[1] for t in tuples}
X_train, y_train = shuffle(train['features'], train['labels'])
X_valid, y_valid = valid['features'], valid['labels']
X_test, y_test = test['features'], test['labels']
Step 1: Dataset Summary & Exploration¶
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 filesignnames.csvcontains id -> name mappings for each id.'sizes'is a list containing tuples, (width, height) representing the 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 IMAGES
Complete 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.
Provide a Basic Summary of the Data Set Using Python, Numpy and/or Pandas¶
### Replace each question mark with the appropriate value.
### Use python, pandas or numpy methods rather than hard coding the results
# Number of training examples
n_train = X_train.shape[0]
# Number of validation examples
n_valid = X_valid.shape[0]
# Number of testing examples.
n_test = X_test.shape[0]
# What's the shape of an traffic sign image?
image_shape = X_train.shape[1:]
# How many unique classes/labels there are in the dataset.
n_classes = len(set(y_train))
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)
Include an exploratory visualization of the dataset¶
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.
### Data exploration visualization code goes here.
# Plot Training / Validation / Test summary counts
for (data, name) in [[y_train, "training"], [y_valid, "validation"], [y_test, "test"]]:
df = pd.DataFrame({'label': data})
counts = df.groupby(['label']).agg({'label': 'count'})
counts.plot(kind='bar', title="counts of each sign in %s data" % name, figsize=(15,4), rot=0)
plt.xlabel("sign")
plt.show()
# Gather 5 Example images per label
examples_per_sign = 5
total = 0
example = {}
for (img,label) in zip(X_train, y_train):
example.setdefault(label, [])
if len(example[label]) < examples_per_sign:
example[label].append(img)
total += 1
if total == n_classes * examples_per_sign:
break;
for label in sorted(example.keys()):
fig = plt.figure()
print(sign_names[label])
for i in range(examples_per_sign):
plt.subplot(1,examples_per_sign,i+1)
plt.imshow(example[label][i])
plt.show()
Step 2: Design and Test a Model Architecture¶
Design and implement a deep learning model that learns to recognize traffic signs. Train and test your model on the German Traffic Sign Dataset.
There are various aspects to consider when thinking about this problem:
- Neural network architecture
- Play around preprocessing techniques (normalization, rgb to grayscale, etc)
- Number of examples per label (some have more than others).
- Generate fake data.
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.
NOTE: 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!
Pre-process the Data Set (normalization, grayscale, etc.)¶
Use the code cell (or multiple code cells, if necessary) to implement the first step of your project.
Model Architecture¶
### Preprocess the data here. Preprocessing steps could include normalization, converting to grayscale, etc.
### Feel free to use as many code cells as needed.
def preprocess(images):
def denoise(img):
"""Denoising did not seem to improve accuracy"""
return cv2.fastNlMeansDenoisingColored(img, h=10)
def normalize(img_yuv):
"""Normalization did not seem to improve accuracy"""
# Global equalization
img = img_yuv.copy()
img[:,:,0] = cv2.equalizeHist(img[:,:,0])
# Local equalization
clahe = cv2.createCLAHE(clipLimit=20.0, tileGridSize=(8,8))
img[:,:,0] = clahe.apply(img[:,:,0])
return img
def convert_to_yuv(img):
"""Conversion into YUV colorspace is effective"""
return cv2.cvtColor(img, cv2.COLOR_RGB2YUV)
return [convert_to_yuv(img) for img in images]
X_train_p = preprocess(X_train)
X_valid_p = preprocess(X_valid)
X_test_p = preprocess(X_test)
# In "Traffic Sign Recognition with Multi-Scale Convolutional Networks", referenced above
# They augmented the dataset by producing 5x permutations on the original training set to
# aid in training
def permute_image(img):
(w, h) = img.shape[:2]
center = (w / 2, h / 2)
rotation = np.random.random()*30-15; # +/- 15 degrees rotation
scale = 1.0+np.random.random()*0.2-0.1 # +/- 10% scaling
M = cv2.getRotationMatrix2D(center, rotation, scale)
rotated = cv2.warpAffine(img, M, (w, h))
return rotated
permutation_factor = 5
X_train_p += [permute_image(img) for img in X_train_p*permutation_factor]
y_train_p = [cls for cls in y_train]
y_train_p += [cls for cls in y_train_p*permutation_factor]
### Define your architecture here.
def conv2d(x, output, stride, name):
weights = tf.Variable(tf.truncated_normal(output), name=name+"_weights")
biases = tf.Variable(tf.zeros(output[3]), name=name+"_biases")
strides = [1, stride, stride, 1]
padding = 'VALID'
return tf.nn.conv2d(x, weights, strides, padding, name=name+"conv") + biases
def maxpool2d(x, k=2, name=""):
return tf.nn.max_pool(x, ksize=[1, k, k, 1], strides=[1, k, k, 1], padding='SAME',
name=name+"_pool")
def fullyconnected(x, output, mu, sigma, name):
weights = tf.Variable(tf.truncated_normal(output, mu, sigma), name=name+"_weights")
biases = tf.Variable(tf.zeros(output[1]), name=name+"_biases")
return tf.add(tf.matmul(x, weights), biases)
def LeNet(x, input_depth=1, n_classes=10):
# Arguments used for tf.truncated_normal, randomly defines variables for the weights and biases for each layer
mu = 0
sigma = 0.1
# Layer 1: Convolutional. Input = 32x32xinput_depth. Output = 28x28x6.
layer1 = conv2d(x, (5,5,input_depth,6), 1, name="layer1")
layer1 = tf.nn.relu(layer1, name="layer1_relu")
layer1 = maxpool2d(layer1, 2, name="layer1")
# Layer 2: Convolutional. Output = 10x10x16.
layer2 = conv2d(layer1, (5,5,6,16), 1, name="layer2")
layer2 = tf.nn.relu(layer2, name="layer2_relu")
layer2 = maxpool2d(layer2, 2, name="layer2")
# Flatten. Input = 5x5x16. Output = 400.
flatten = tf.contrib.layers.flatten(layer2)
# Layer 3: Fully Connected. Input = 400. Output = 120.
layer3 = fullyconnected(flatten, [400, 120], mu, sigma, name="layer3")
layer3 = tf.nn.relu(layer3, name="layer3_relu")
# Layer 4: Fully Connected. Input = 120. Output = 84.
layer4 = fullyconnected(layer3, [120,84], mu, sigma, name="layer4")
layer4 = tf.nn.relu(layer4, name="layer4_relu")
#Layer 5: Fully Connected. Input = 84. Output = n_classes.
logits = fullyconnected(layer4, [84, n_classes], mu, sigma, name="layer5")
return logits
Train, Validate and Test the Model¶
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 test set but low accuracy on the validation set implies overfitting.
### 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.
rate = 0.00001
input_depth = image_shape[2]
x = tf.placeholder(tf.float32, (None, 32, 32, input_depth), name="X")
y = tf.placeholder(tf.int32, (None), name="y")
one_hot_y = tf.one_hot(y, n_classes)
logits = LeNet(x, input_depth, n_classes)
cross_entropy = tf.nn.softmax_cross_entropy_with_logits(logits=logits, labels=one_hot_y)
loss_operation = tf.reduce_mean(cross_entropy)
optimizer = tf.train.AdamOptimizer(learning_rate = rate, name="optimizer")
training_operation = optimizer.minimize(loss_operation)
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))
def evaluate(sess, X_data, y_data):
num_examples = len(X_data)
total_accuracy = 0
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})
total_accuracy += (accuracy * len(batch_x))
return total_accuracy / num_examples
saver = tf.train.Saver()
sess = tf.Session()
do_restore = True
try:
if do_restore:
print("restoring from save file")
saver.restore(sess, './lenet_signs')
else:
print("initializing variables")
sess.run(tf.global_variables_initializer())
except Exception as ex:
print(ex)
print("initializing variables")
sess.run(tf.global_variables_initializer())
rate = 0.0001
EPOCHS = 100
BATCH_SIZE = 128
num_examples = len(X_train_p)
print("Training...")
print()
for i in range(EPOCHS):
for offset in range(0, num_examples, BATCH_SIZE):
# Get the batch
end = offset + BATCH_SIZE
batch_x, batch_y = X_train_p[offset:end], y_train_p[offset:end]
# Run the training operation
sess.run(training_operation, feed_dict={x: batch_x, y: batch_y})
validation_accuracy = evaluate(sess, X_valid_p, y_valid)
print("EPOCH {} ...".format(i+1))
print("Validation Accuracy = {:.3f}".format(validation_accuracy))
print()
saver.save(sess, './lenet_signs')
print("Model saved")
# Accuracy on the training set
# Not stastically useful for generalization, however if we have near-perfect accuracy
# here, but not on validation and test data that is a signal that we have likely overfit
# the training data
train_accuracy = evaluate(sess, X_train_p, y_train_p)
print("Train Accuracy = {:.3f}".format(train_accuracy))
# Evaluate Validation accuracy
valid_accuracy = evaluate(sess, X_valid_p, y_valid)
print("Validation Accuracy = {:.3f}".format(valid_accuracy))
# Evaluate Test accuracy
test_accuracy = evaluate(sess, X_test_p, y_test)
print("Test Accuracy = {:.3f}".format(test_accuracy))
### More on accuracy, display a confusion matrix of the test results
prediction = sess.run(tf.argmax(logits, 1), feed_dict={x: X_test_p})
cnf_matrix = confusion_matrix(y_test, prediction)
# Normalize
cnf_matrix = cnf_matrix.astype('float') / cnf_matrix.sum(axis=1)[:, np.newaxis]
# Hack to get the matrix to display nicely by encourage all values to be 1 digit integers
cnf_display = (cnf_matrix*10).astype('int')
np.set_printoptions(threshold=2000, linewidth=500)
print(cnf_display)
np.set_printoptions()
## Take a closer look at some signs that we did poorly on (< 40% of results correct)
df = pd.DataFrame({'label': y_train})
counts = np.array(df.groupby(['label']).agg({'label': 'count'}))
diag = np.diagonal(cnf_matrix)
poor = [i for (i, v) in enumerate(diag) if v < 0.40]
for offset in poor:
print(sign_names[offset], "-", counts[offset], "examples in training set")
classes = [i for (i, v) in enumerate(cnf_matrix[offset]) if v > 0.05]
for index in classes:
print(" ", "%4.1f%% -" % (cnf_matrix[offset][index]*100), sign_names[index])
Analysis of the above:
All of the worst performing sign classes had few examples in the training set which leads both to a low apriori expectation for those signs as well as insufficient training examples to be able to differentiate them well.
For example "Speed limit (20km/h)" had only 180 examples in the training set where there were > 1000 examples of all the other speed limit signs, and ~2000 examples on both "Speed limit (30km/h)" and "Speed limit (50km/h)". With close to a 10:1 apriori bias torwards "Speed limit 30km/h" over "Speed limit (20 km/h)", it should not be completely unexpected that the misclassification results tend to favor classification as 30/km over 20km/h.
This overall trend continues with the other signs with signicant misclassification error where all of them have relatively few examples in the training data compared to the signs with better classification rates.
The primary remedy for this would be to increase the amount of training data for the classes of signs that are currently significantly under-represented in the training data.
Step 3: Test a Model on New Images¶
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.
Load and Output the Images¶
### Load the images and plot them here.
import matplotlib.image as mpimg
# Load new images, all images have been pre-cropped to square images
mgdir = 'new_images'
X_new_images = []
for filename in os.listdir(imgdir):
if filename.endswith('.jpg'):
img = mpimg.imread(os.path.join(imgdir, filename))
img = cv2.resize(img, (32,32), interpolation=cv2.INTER_AREA)
X_new_images.append(img)
# Plot the new images
vis = np.concatenate(X_new_images[0:6], axis=1)
plt.imshow(vis)
plt.show()
vis = np.concatenate(X_new_images[6:], axis=1)
plt.imshow(vis)
plt.show()
Predict the Sign Type for Each Image¶
### 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.
X_new_images_p = preprocess(X_new_images)
y_new_images = [33, 17, 27, 3, 2, 14, 11, 18, 13, 28, 38, 40]
prediction = sess.run(tf.argmax(logits, 1), feed_dict={x: X_new_images_p})
for (i,v) in enumerate(prediction):
annotation = "Actual: %s\nPredicted: %s" % (
sign_names[y_new_images[i]], sign_names[v])
fig = plt.figure(figsize=(1,1))
plt.imshow(X_new_images[i])
plt.annotate(annotation,xy=(0,0), xytext=(60,25), fontsize=12, family='monospace')
plt.show()
Analyze Performance¶
### Calculate the accuracy for these 5 new images.
### For example, if the model predicted 1 out of 5 signs correctly, it's 20% accurate on these new images.
accuracy = sum(y_new_images == prediction)/len(y_new_images)
print("Accuracy = %4.2f%%" % (100*accuracy))
Output Top 5 Softmax Probabilities For Each Image Found on the Web¶
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.
### Print out the top five softmax probabilities for the predictions on the German traffic sign images found on the web.
k = 5
top_k = tf.nn.top_k(tf.nn.softmax(logits), k=k)
top = sess.run(top_k, feed_dict={x: X_new_images_p})
#print(top[0])
for i in range(len(y_new_images)):
print("\n == Actual == -", sign_names[y_new_images[i]])
for j in range(k):
print(" %22.20f%% - %s" % (top[0][i][j], sign_names[top[1][i][j]]))
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. 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.