Autonomous Vehicle System

The Project

This image classification model was made in Google Colab using the Python TensorFlow library. The training dataset consisted of around 900 images, and the model is able to recognize and classify the stop sign, crosswalk sign, traffic light, and speed limit sign. I chose these four signs because they're the most crucial ones to look out for when operating a car on the road. Once the model receives graphical input (a photo) from the user's camera, it is able to give a prediction to the road sign that's ahead. In training the model, an accuracy of 92% was achieved, with only 7 out of 88 images identified incorrectly. The project is not finished, as I'm hoping to learn more about the operations of onboard computers of autonomous vehicles. I hope to be able to implement my model into a car as a trial!

View this project on YouTube

Team Comments

I began the project when I had been heading home from school, and had seen someone almost run a pedestrian over because he had been distracted on his phone. Because so many fatal automotive accidents are caused by human error, building an autonomous system could aid in reducing loss of life.

I found debugging the program to be most difficult. By the time I had gotten halfway through the training of the model, I was reciting the TensorFlow documentation in my sleep! I learned that it helps very much to be experienced with the libraries you wish to implement before beginning to work.

Of course! I would like to add a feature that identifies pedestrians ahead, plus a Voiceover-like feature that announces to the driver when there are pedestrians in the lane. It could be a good safety net to the driver in case he/she did not spot the pedestrians.