VRIC: Vehicle Re-Identificaton in Context

VRIC

VeRi-776

VehicleID

Description

Existing vehicle re-identification (re-id) evaluation benchmarks consider strongly artificial test scenarios by assuming the availability of high quality images and fine-grained appearance at an almost constant image scale, reminiscent to images required for Automatic Number Plate Recognition, e.g. VeRi-776. Such assumptions are often invalid in realistic vehicle re-id scenarios where arbitrarily changing image resolutions (scales) are the norm. This makes the existing vehicle re-id benchmarks limited for testing the true performance of a re-id method. In this work, we introduce a more realistic and challenging vehicle re-id benchmark, called Vehicle Re-Identification in Context (VRIC). In contrast to existing datasets, VRIC is uniquely characterised by vehicle images subject to more realistic and unconstrained variations in resolution (scale), motion blur, illumination, occlusion, and viewpoint. It contains 60,430 images of 5,622 vehicle identities captured by 60 different cameras at heterogeneous road traffic scenes in both day-time and night-time.

Download

VRIC Dataset (129MB): [Google Drive]

Citation

            Vehicle Re-Identificaiton in Context.
            Aytac Kanaci, Xiatian Zhu and Shaogang Gong.
            Pattern Recognition - 40th German Conference, (GCPR) 2018, Stuttgart
            Paper 
             
            

            @inproceedings{2018gcpr-Kanaci,
            author    = {Aytac Kanaci and
                         Xiatian Zhu and
                         Shaogang Gong},
            title     = {Vehicle Re-Identification in Context},
            booktitle = {Pattern Recognition - 40th German Conference, {GCPR} 2018, Stuttgart,
                         Germany, September 10-12, 2018, Proceedings},
            year      = {2018}
            }
            

Related Datasets

VRIC is a result of a conversion of from UA-DETRAC vehicle tracking dataset. A conversion using the ground truth tracking identities were used in the creating of the re-identification labels. More details can be found in our paper.

Licence

All the images were collected from UA-DETRAC and the copyright belongs to the original owners.

@article{DETRAC:CoRR:WenDCLCQLYL15,
  author    = {Longyin Wen and Dawei Du and Zhaowei Cai and Zhen Lei and Ming{-}Ching Chang and
               Honggang Qi and Jongwoo Lim and Ming{-}Hsuan Yang and Siwei Lyu},
  title     = { {UA-DETRAC:} {A} New Benchmark and Protocol for Multi-Object Detection and Tracking},
  journal   = {arXiv CoRR},
  volume    = {abs/1511.04136},
  year      = {2015}
}          

Contact

Please feel free to send any questions, comments, and evaluation results with a brief method description to Aytac Kanaci at a.kanaci@qmul.ac.uk.