UQTR Dataset - Snowy and non-snowy road images

Citation Author(s):
Tatiana
Ortegon-Sarmiento
Université du Québec à Trois-Rivières, Universidad de Granada
Sousso
Kelouwani
Université du Québec à Trois-Rivières
Ali
Amamou
Université du Québec à Trois-Rivières
Jonathan
Boisclair
Université du Québec à Trois-Rivières
Alvaro
Uribe-Quevedo
Ontario Tech University
Muhammad Zeshan
Alam
Brandon University
Patricia
Paderewski-Rodriguez
Universidad de Granada
Francisco
Gutierrez-Vela
Universidad de Granada
Submitted by:
Tatiana Ortegon...
Last updated:
Mon, 03/24/2025 - 18:40
DOI:
10.21227/q4af-r432
Data Format:
Links:
License:
0
0 ratings - Please login to submit your rating.

Abstract 

 

The UQTR dataset consists of 7838 real and synthetic images of the Université du Québec à Trois-Rivières (UQTR) campus road under normal and snow conditions. The image resolution is 1280×720. It includes lane labels in .txt files, where each row stores the set of points of a lane. The points are stored as x1 y1 x2 y2, as in the tutorial by Ruijin Liu, Zejian Yuan, Tie Liu, Zhiliang Xiong: Train and Test Your Custom Data.

Synthetic data were generated using different image synthesis techniques from real road data to improve lane detection in snowy winter conditions. Simulation data, corresponding to a virtual environment of the UQTR campus, are also included.

The dataset includes:

  • 967 real road images
  • 6655 images with synthetic snow
  • 216 images of the virtual environment

References to 3D models of the virtual environment:

  1. Feel funny, “Shelter a00,” 2010, accessed July 2023. [Online]. Available: https://www.turbosquid.com/3d-models/free-max-model-bus-shelter/520824 
  2. NinjRosa, "Stop," 2022, accessed July 2023. [Online]. Available: https://www.turbosquid.com/3d-models/stop-1958611 
  3. Adobe Systems Incorporated, "Mixamo Bryce character," 2015, accessed July 2023. [Online]. Available: https://www.mixamo.com/#/?page=2&type=Character

 

 

Instructions: 

This dataset can be used for training and testing deep learning models of lane detection. It consists of images, in .jpg format, and lane labels, in .txt files, and is divided into two datasets, a training dataset, with 6271 images, and a test dataset, with 1567 images. The images and their respective labels, whose names are the same but with different formats, are distributed in separate folders for ease of use.

To use this datset:

  1. Download the dataset.
  2. Load images from train_images and test_images folders.
  3. Load labels from train_labels and test_labels folders.
  4. Model training
Funding Agency: 
Fonds de Recherche du Québec -Nature et Technologies, the Canada Research Chair program, the NSERC, and the Ministry of Science and Innovation of Spain (MCIN/AEI/ 10.13039/501100011033, UE)
Grant Number: 
CRC-950-232172 (CRC-2018-00299), Discovery Grants RGPIN-2018-05917, CRSNG-RGPIN-2019- 07206, and CRSNG-RGPAS-2019-00114, and the PLEISAR-Social project (PID2022-136779OB-C33)

Comments

Free Chat with Astrologer to access live Astrology Consultation 24x7 and solutions to your worries. Get the First free astrology chat with the Best Astrologer. Visit: https://astroera.in/chat-with-astrologer

Submitted by astroera era on Wed, 03/26/2025 - 09:22

Documentation

AttachmentSize
File Description of the dataset structure395.99 KB