Skip to content

Latest commit

 

History

History
59 lines (45 loc) · 3.56 KB

README.md

File metadata and controls

59 lines (45 loc) · 3.56 KB
Raw

DeepGTAV

ScriptHook mod to build a vision-based deep learning environment using GTAV. It supports two main operation modes, Dataset generation or Reinforcement Learning environment (under development)

Self-Driving Car

Installation instructions

Update GTAV to the last version and just copy-paste the contents of bin/Release under your GTAV installation directory. Then set the correct dataset destination directory in the config.ini file and run the game (story mode).

It will work out-of-the-box, it doesn't require any extra dependencies!

Configuration options

The configuration options of the environment are set through the config.ini file:

  • mode: 0 for dataset generation, 1 for reinforcement learning (TODO) [integer]
  • imageWidth: Width in pixels of the images to store during dataset generation [integer]
  • imageHeight: Height in pixels of the images to store during dataset generation [integer]
  • captureFreq: Frequency of image and labels capture in Hz (FPS) [integer]
  • datasetDir: Absolute path with trailing slash to directory to save the dataset (must exist) [string]
  • maxDuration: Duration in hours of the dataset generation [integer]
  • setSpeed: Set speed of the vehicle in m/s during dataset generation [float]
  • drivingStyle: Driving style of the vehicle driver during dataset generation [integer]
  • car: Vehicle to use during dataset generation. For now only blista (Toyota Prius) is supported (0) [integer]
  • weatherChangeDelay: Time in seconds between weather changes [integer]
  • initialWeather: Initial weather type [integer]
  • initialHour: Initial in-game hour [integer]
  • initialMinute: Initial in-game minute [integer]
  • initialPosX: Initial x position of the vehicle (will be redefined to closest road) [integer]
  • initialPosY: Initial y position of the vehicle (will be redefined to closest road) [integer]

See Scenario.cpp for more details on the supported values, especially for car, drivingStyle and initialWeather.

Dataset generation description

The in-game screenshots are stored as RGB PNG format, with the specified width and length of the config.ini file. These images are named in order of capture from 1 to undefined. Alongside the images, a file named dataset.txt contains a new line for each image name, with the labels associated to it separated by spaces. The labels are the following in the same order:

  • Speed (m/s)
  • Acceleration (m/s2)
  • Brake pedal position (0 to 1)
  • Steering angle (-1 to 1, left to right)
  • Throttle pedal position (0 to 1)
  • Yaw rate (deg/s)
  • Direction (-1 to left, 1 to right)

This way of storing the dataset has two main benefits:

  • It allows easy shuffling and splitting of data, by just reordering the lines or creating new dataset files.
  • Easy preview and/or modification of the images

It can also have some drawbacks:

  • Generates a lot of files, which can be hard to handle by graphical file browsers
  • Higher size, given by the generation of meta-data produced by the individual files.

TODO

  • Add more cars
  • Find a way to automatically close/restart the game once dataset generation has finished/vehicle is stuck.
  • Implement RL mode

Acknowledgements

Many thanks to @crizCraig and the OpenAI team for providing me with access to their repos and pioneering this work with DeepDrive!