Self-driving car

Developing a «mind» for self-driving car prototype.

Deep learning
Computer vision
CUDA
ConvNets

2017

About project

Task

Develop a self-driving car software prototype, capable of driving in any weather conditions within business center parking lot.

Solution

  • Scene recognition module based on conv nets
  • Control module — driving direction selection
  • Data module — datasets collection and augumentation
  • Stereo vision module — Deep Learning based software that calculates distance from video

Autonomous driving is here

There’s no doubt that self-driving cars are the future. Today autopilots is a hot topic. Every major automaker attempts to create their own version of an autonomous or semi-autonomous driving. We decided to join the hype train.

Together with company Bright Box we made a self-driving car prototype, capable of driving in any weather conditions within business center parking lot.

Below you will find a video of test runs with autonomous driving. Pay attention, that driver doesn't touch the steering wheel. Every movement is based on video input from camera, mounted on the car roof.

Video 1. Test self-driving car run.

Video 2. Test self-driving car run.
Next project

Autonomous forklift