ynoza/Traffic

This project creates a model using neural network layers that can predict which street sign is which. To do this, the project took use of the German Traffic Sign Recognition Benchmark, which contains a huge data base of 43 common street signs. This data was then organized into numpy arrays, and then placed into training and testing sets. A neural network model was then created using many hidden layers (i.e convolutional and average pooling) to train the model on our data set. The model was then finally evaluated and an accuracy percentage of about 97.5% was recorded. Huge credits go to Harvard's CS_50 course, in particular the introduciton to artificial intelligence with python portion of the course. This repository is a project that was assigned to me from this course, so CS_50 has ownership to portions of the source code.

Traffic

This project creates a model using neural network layers that can predict which street sign is which. To do this, the project took use of the German Traffic Sign Recognition Benchmark, which contains a huge data base of 43 common street signs. This data was then organized into numpy arrays, and then placed into training and testing sets. A neural network model was then created using many hidden layers (i.e convolutional and average pooling) to train the model on our data set. The model was then finally evaluated and an accuracy percentage of about 97.5% was recorded. Huge credits go to Harvard's CS_50 course, in particular the introduciton to artificial intelligence with python portion of the course.

Instructions:

First unzip the folder.

Then run python3 traffic.py, and enjoy :).

View:

The algorithm goes over more than 150000 images (each Epoch is unique), and as you can see the model trains itself and gets better each time.