Vehicle search in the Wild: A New Benchmark
-
This repository contains the VSW dataset for the following paper：Unsupervised Vehicle search in the Wild: A New Benchmark (ACM MM 2021).
-
Vehicle search is a crucial part of the city's surveillance system. However, existing research does not fully explore end-to-end vehicle search. Instead, VSW aims to advance the task of vehicle search and benefit traffic regulation and criminal investigations.
-
Here https://pan.baidu.com/s/14aKlzkMObz33iAI6WwSYpQ (Extraction code: uk0n). This dataset is for research purposes only and therefore not for commercial use.
-
If you have any questions about the VSW dataset, please contact zhongx@whut.edu.cn.
-
VSW for vehicle search.
-
Note that, we publish only part of the data, also for the initial version. It contains 76,387 frames and 475 ids. We split the dataset into a train and a test subset, ensuring no overlapped images or labeled identities between them. The train set contains 38,280 frames and 254 training IDs, the first 99 are cross-camera ids. The test set contains 38,280 frames and 254 training IDs, the first 97 are cross-camera ids. Then for each vehicle in the test set, we randomly select two bounding boxes as query vehicles under each camera.
- This dataset is organized as follows:
├── annotations # Annotation of frames
│ ├── */*.mat
├── frames # the vehicle frames
│ ├── */*.jpg
├── query # query set
│ ├── */*.jpg
├── frame_train.mat # frames for model training
├── frame_test.mat # frames for model testing
├── ID_train.mat # IDs for model training
├── ID_test.mat # IDs for model testing
├── query_info.txt # query set
If you find our dataset useful in your research work, please consider citing:
bib:
@inbook{10.1145/3474085.3475654,
title = {Unsupervised Vehicle Search in the Wild: A New Benchmark},
author = {Zhong, Xian and Zhao, Shilei and Wang, Xiao and Jiang, Kui and Liu, Wenxuan and Huang, Wenxin and Wang, Zheng},
booktitle = {Proceedings of the 29th ACM International Conference on Multimedia},
year = {2021},
pages = {5316–5325},
}