Project: Build a Traffic Sign Recognition Program

Overview

In this project, we will use what we've learned about deep neural networks and convolutional neural networks to classify traffic signs. We will train and validate a model so it can classify traffic sign images using the German Traffic Sign Dataset. After the model is trained, we will then try out our model on images of German traffic signs that we find on the web.

The Project

The goals / steps of this project are the following:

• Load the data set
• Explore, summarize and visualize the data set
• Design, train and test a model architecture
• Use the model to make predictions on new images
• Analyze the softmax probabilities of the new images
• Summarize the results with a written report

Dependencies

This lab requires:

• CarND Term1 Starter Kit

The lab environment can be created with CarND Term1 Starter Kit. Click here for the details.

Dataset and Repository

1. Download the data set. The classroom has a link to the data set in the "Project Instructions" content. This is a pickled dataset in which we've already resized the images to 32x32. It contains a training, validation and test set. Dataset can also be downloaded from here
2. Clone the project, which contains the Ipython notebook and the writeup template.

git clone https://github.com/udacity/CarND-Traffic-Sign-Classifier-Project
cd CarND-Traffic-Sign-Classifier-Project
jupyter notebook Traffic_Sign_Classifier.ipynb

Implementation

Implementation is provided in Traffic_Sign_Classification.ipynb

• Acheived test accuracy of 93.55 %
• Adam Optimizer is used. The tf.train.AdamOptimizer uses Kingma and Ba's Adam algorithm to control the learning rate. Adam offers several advantages over the simple tf.train.GradientDescentOptimizer. Foremost is that it uses moving averages of the parameters (momentum). Simply put, this enables Adam to use a larger effective step size, and the algorithm will converge to this step size without fine tuning. The main down side of the algorithm is that Adam requires more computation to be performed for each parameter in each training step (to maintain the moving averages and variance, and calculate the scaled gradient); and more state to be retained for each parameter (approximately tripling the size of the model to store the average and variance for each parameter). A simple tf.train.GradientDescentOptimizer could equally be used, but would require more hyperparameter tuning before it would converge as quickly.
• Batch size was set to 50
• Model is to be trained for 200 epochs but can stop training earlier if validation accuracy for all the batches in validation dataset is greater then or equal to 98%.

Please, have a look on the code for detailed implementation.