Arvutiteaduse instituut
  1. Esileht
  2. ADL Minicar Challenge
EN
Logi sisse

ADL Minicar Challenge

  • Main
  • 2021

Data Recording
Windows v. Linux
Example

  • 2022
  • The Car
  • Projects with DonkeyCar

DATA and DATA COLLECTION

The public dataset can be found here:


(link to data)

Data is organized by recordings. The most important data for each recording are the frames and the JSON file that contains timestamps for these frames and the timestamped commands given to the car by the human driver. To better understand the data format check the Documentation from RCSnail.

The data is cleaned - despite there being collisions in the video files, these sections of the sessions are removed from the frames and JSON files.

How the visual inputs look like, can be seen in the video below (human driving, also shows camera view):
https://drive.google.com/file/d/1ccim7tzmPKkdyU-s4siVsu7TUhcdjSVU/view?usp=sharing

You can record your own data. Currently, this can be done only in Windows and only using either keyboard or xbox controller. In Linux data recording is implemented in AI-lite, but this records already downsized frames, not the original 1280x640 frames. If that is fine for you, you can try recording that way.

IMPORTANT when using RCSnail-AI-lite to control the car, pay attention that the frames it receives are already downscaled, not the original size as the public dataset. Your model performance might depend on it. The default downsizing method is PIL.Image.NEAREST. It can be changed to linear scaling in the config file (ask for details, if needed).

Read about the commands that need to be sent to the car in Windows v. Linux, as this is different for different operating systems.


Recording at Delta track

Track is located on the 2nd floor student area.

  • You need to connect to the rcsxg wifi network to connect to the car. If that does not work, you can also attempt from eduroam, but it's less likely to succeed and slower (not direct connection trhough the hotspot on site)
  • The car also takes up to a minute to start in the first place. Always make sure your car's battery is not dead (check if processor lights are visile inside the car, voltage displayed in user interface should be 3400mV or more). Always keep the extra batteries recharging while using the car!
  • The car disconnects sometimes. It is a common bug that the car disconnects at 10 seconds into first session after being restarted. You connect to the car via launching the RCSnail.exe and clicking "New session" or by launching src/main.py in RCSnail-Connector. A connection is created and you will see camera feed on screen. But the first connection disconnects shortly and the camera feed freezes. Close session (via button in UI) or closing the RCNsail-Connector and relaunch. The second connection will be rather stable. If it eventually disconnects at a random moment, usually after a few minutes, again don't restart the car, just the connection.
  • Arvutiteaduse instituut
  • Loodus- ja täppisteaduste valdkond
  • Tartu Ülikool
Tehniliste probleemide või küsimuste korral kirjuta:

Kursuse sisu ja korralduslike küsimustega pöörduge kursuse korraldajate poole.
Õppematerjalide varalised autoriõigused kuuluvad Tartu Ülikoolile. Õppematerjalide kasutamine on lubatud autoriõiguse seaduses ettenähtud teose vaba kasutamise eesmärkidel ja tingimustel. Õppematerjalide kasutamisel on kasutaja kohustatud viitama õppematerjalide autorile.
Õppematerjalide kasutamine muudel eesmärkidel on lubatud ainult Tartu Ülikooli eelneval kirjalikul nõusolekul.
Courses’i keskkonna kasutustingimused