UFPR‑VCR Dataset

Overview

UFPR‑VCR (UFPR Vehicle Color Recognition) dataset aims to address more complex vehicle color recognition scenarios than previous studies. The dataset’s construction and preliminary deep‑learning experiments are detailed in our paper Toward Enhancing Vehicle Color Recognition in Adverse Conditions: A Dataset and Benchmark.

Dataset Details

• Number of Images: 10,039
• Number of Vehicles: 9,502
• Vehicle Types: Cars, trucks, buses, and vans
• Image Conditions: Front and rear views, partial occlusions, varied illumination, and nighttime scenes
• Color Classes: 11 colors (beige, black, blue, brown, gray, green, orange, red, silver, white, yellow)
• Source: Collected from six public datasets in Brazil originally for Automatic License Plate Recognition (ALPR)
• Processing: Original images were pre‑processed and filtered to standardise them and to retain those suitable for color recognition
• Annotation Validation: Over 90 % of vehicle annotations were verified using licence‑plate information from the original datasets

Access

Dataset access requires signing a license agreement and is provided free of charge for non‑commercial academic research.

Citation

When using the UFPR‑VCR dataset, please cite our paper:

@inproceedings{lima2024toward,
  title = {Toward Enhancing Vehicle Color Recognition in Adverse Conditions: A Dataset and Benchmar},
  author = {G. E. {Lima} and R. {Laroca} and E. {Santos} and E. {Nascimento Jr.} and D. {Menotti}},
  year = {2024},
  month = {Sept},
  booktitle = {Conference on Graphics, Patterns and Images (SIBGRAPI)},
}

Contact

For questions or comments, please contact Gabriel E. Lima (gelima@inf.ufpr.br).