Mimic

Neural networks have been advancing in capability very rapidly in recent years. One of the newest techniques with these networks is Generative Adversarial Networks. In this GAN architecture you have two neural networks pitted against each other, one trying to fool the other with noise, while the other trains on real data and responds with information on how to make that noise more realistic. After many runs, you would ideally be able to generate data that the other network wouldn't know was real or fake. Currently autonomous vehicles require a lot of training data. This data can be very expensive to get and label. We propose using the generator in the GAN to create synthetic data from real traffic sign images that have already been collected and labeled. These images would not have to be labeled and would ideally be accurate yet different representations of the real data. We hypothesize that training a model on the real + synthetic data we generate will lead to higher accuracy in image classification, without the need to go through the painful process of collecting and labeling new and real data.

Built With

• jupyter
• python
• tensorflow