Traffic Sign Classification/ Recognition

---

Objectives

The objectives of this project are:

• Implement deep neural networks and convolutional neural networks to classify traffic signs.
• Classify the German Traffic Sign Dataset using the post popular and highly used classification architecture called LeNet.
• Design, train and test a model architecture
• Use the model to make predictions on new images
• Analyze the softmax probabilities of the new images

---

Dataset

For this project, I used the German Traffic Sign dataset for classification. Following table illustrates the dataset.

Description	Value
The size of training set is	34799
The size of the validation set is	4410
The size of test set is	12630
The shape of a traffic sign image is	(32, 32, 3)
The number of unique classes/labels in the data set is	43

Following are randomly picked images with their labels each from a different class.

Training Set

The figure below illustrates number of image samples per class in the training set.

Validation Set

The figure below illustrates number of image samples per class in the validation set.

Testing Set

The figure below illustrates number of image samples per class in the testing set.

Image Pre-processing

The image is first normalized to have pixel values between -1.0 to 1.0 and also have zero mean. The images are trained relatively faster with the normalization. Further, the image is converted to a grayscale color space from RGB color space. This reduces the breadth of the layering. The images are trained in batches and each batch consists of images shuffled to eliminate any bias.

Model Architecture

Design

LeNet is a popular classification architecture for digits, traffic signs, etc. My design consists of layers as tabulated below.

Layer	Description	Input	Output
Input	32x32x1 normalized gray scale image	32x32x1	32x32x1
Convolution 3x3	1x1 stride, VALID padding, activation RELU	32x32x1	28x28x48
Max pooling	2x2 stride, 2x2 kernel	28x28x48	14x14x48
Convolution 3x3	1x1 stride, VALID padding, activation RELU	14x14x48	10x10x96
Max pooling	2x2 stride, 2x2 kernel	10x10x96	5x5x96
Convolution 3x3	1x1 stride, VALID padding, activation RELU	5x5x96	3x3x172
Max pooling	2x2 stride, 2x2 kernel	3x3x172	2x2x172
Flatten	flattens the the 2D layer to a vector	2x2x172	688
Fully connected	Artificial NN with activation RELU	688	84
Fully connected	Artificial NN with softmax probabilities	84	43

Training and Validation

As mentioned earlier the images are trained in batches. EPOCHS or episodes are run with asingle batch trained in it. Following are the parameters used for training.

EPOCHS = 17 ................. After running 17 epochs there is no significant or no improvement in the accuracy. BATCH_SIZE = 128 ................. I trained the network model on a local CPU and hence preferred a low batch size of 128 images per batch. LEARNING RATE = 0.001 ................. Since Adam optimizer was used, a learning rate of 0.001 is suggested.

Following are the accuracies each for the training set, the last validation set and the testing set.

Set	Value
Training set	0.997
Validation set	0.938
Test set	0.929

The graph below shows a trade off between the training and validation accuracies considering the number of episodes run.

Testing on Unknown Images

I picked the following 5 unknown images for testing.

The classifier is able to correctly classify 4 out of 5 images yielding an accuracy odf 80.00%. The classifier incorrectly classifies the 4th image as 'General caution' instead of 'Pedestrians'. If we notice individual probabilties for the 4th image, the probabilities of the image being 'General caution' and 'Roundabout' are close. Although, the image is taken in perfect lighting conditions, the classifier is confused between an exclaimaton '!' in 'General caution' and the stick figure of a pedestrianin 'Pedestrians'.

The prediction for the images are tabulated as follows :

Image	Prediction
Keep right	Keep right
Stop	Stop
Speed limit (30km/h)	Speed limit (30km/h)
Pedestrians	General caution
No entry	No entry

The probabilities for individual labels for each image are as follows :

1. Image 1 - Keep Right

Labels	Probability
Keep right	1.00
Speed limit (50km/h)	0.00
Yield	0.00
Speed limit (30km/h)	0.00
Priority road	0.00

2. Image 2 - Stop

Labels	Probability
Stop	1.00
Go straight or right	0.00
Road work	0.00
Turn right ahead	0.00
Priority road	0.00

3. Image 3 - Speed limit (30km/h)

Labels	Probability
Speed limit (30km/h)	1.00
Road work	0.00 %
Speed limit (20km/h)	0.00
Speed limit (50km/h)	0.00
Speed limit (80km/h)	0.00

4. Image 4 - Pedestrians

Labels	Probability
General caution	0.55
Roundabout mandatory	0.45
Pedestrians	0.00
End of speed limit (80km/h)	0.00
Priority road	0.00

4. Image 5 - No entry

Labels	Probability
No entry	1.00
Stop	0.00 %
Turn right ahead	0.00
Turn left ahead	0.00
No passing	0.00