DataGAN: Leveraging Synthetic Data for Self-Driving Vehicles

The boxing match: generator vs discriminator

  1. The generator: this is the model that creates your generative outputs. By taking in latent (random) noise, the generator is then able to upsample it to create realistic images.
  2. The discriminator: the main goal behind the discriminator is to determine whether the image, created by the generator, is real or fake. This is where the term “adversarial” comes from in GANs → the generator and discriminator compete to see whether the generator can create images that can fool the discriminator.
An example of how GANs work. Source
You want both generator and discriminator to be even in strength!

Deep Convolutional Generative Adversarial Networks (DCGANs)

Original proposal for DCGANs implementation. Source
  • Both networks (discriminator + generator) are Fully Convolutional Networks (FCN). What this means is that there are no Dense/maxpooling layers after convolutions.
  • ReLU is used as the activation function for generators while LeakyReLU is used for the discriminator
  • Batch normalization is now a method that’s consistently used to help with gradient flow and avoid vanishing/exploding gradient problems
  • Tanh is the output function and should be used instead of something like sigmoid (normalization between [-1, 1] instead of [0, 1])

DataGAN

Source

Current Progress

Why I built DataGAN

More updates soon on Github.

Next steps

  • Build a straightforward lane detection model, fully trained on DataGAN outputs that are functional
  • Create the first self-driving GAN dataset that can be used to train robust lane detection + computer vision models
  • Make an AC-GAN model where you could potentially input an edge case scenario (ex. overtaking a vehicle) and use GAN-ConvLSTM to help generate the given scene

--

--

--

Building Self-Driving Cars as a 15yo. srianumakonda.com

Love podcasts or audiobooks? Learn on the go with our new app.

Recommended from Medium

Model & Variable Selection Methods!

Decipher Text Insights and Related Business Use Cases

Recommender Systems: Exploring the Unknown Using Uncertainty

Paper Review: Elastic Averaging SGD and Consensus-based Distributed SGD for Distributed Training of…

This (StackOverflow) Question Does Not Exist.

Uncertainty for CTR Prediction: One Model to Clarify Them All

Product Resale Price Prediction using Machine Learning — a Case Study

I personally really enjoyed the reading of this paper as it helped encourage thought experiments…

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Sri Anumakonda

Sri Anumakonda

Building Self-Driving Cars as a 15yo. srianumakonda.com

More from Medium

Representation Learning: What and Why

Training a CNN to detect Face Masks

How does AI impact self-driving cars?

Understanding Attention in Machine Learning