ShanghaiTech Campus dataset (Anomaly Detection)

ShanghaiTech University

[Download (Onedrive)]

Introduction

It is desirable that the anomaly detection model trained can be directly applied in multiple scenes with multiple view angles. However, almost all existing datasets only contain videos captured with one fixed angle camera, and it lacks diversity of scenes and view angles. To increase scene diversity, we build a new anomaly detection dataset. Further we introduce the anomalies caused by sudden motion in this dataset, such as chasing and brawling in our dataset, which are not included in existing datasets. These characteristics make our dataset more suitable in real scenarios. To better understand the differences between our dataset and existing anomaly detection datasets, we briefly summarize all anomaly detection datasets as follows:

CUHK Avenue dataset [1] contains 16 training videos and 21 testing videos with a total of 47 abnormal events, including throwing objects, loitering and running. The size of people may change because of the camera position and angle.

Pedestrian 1 (Ped1) dataset [2] includes 34 training videos and 36 testing videos with 40 irregular events. All of these abnormal cases are about vehicles such as bicycles and cars. Pedestrian 2 (Ped2) [2] dataset contains 16 training videos and 12 testing videos with 12 abnormal events. The definition of anomaly for Ped2 is the same with Ped1.

Subway dataset [3] are 2 hours long in total. There are two categories, i.e. Entrance and Exit. Unusual events contain walking in wrong directions and loitering. More importantly, this dataset is recorded in indoor environment while above ones are recorded in outdoor environment.

>

ShanghaiTech Campus

ShanghaiTech Campus dataset has 13 scenes with complex light conditions and camera angles. It contains 130 abnormal events and over 270, 000 training frames. Moreover, pixel level ground truth of abnormal events is also annotated in our dataset.

Citation

If you find this useful, please cite our work as follows:
@INPROCEEDINGS{liu2018ano_pred, 
        author={W. Liu and W. Luo, D. Lian and S. Gao}, 
        booktitle={2018 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)}, 
        title={Future Frame Prediction for Anomaly Detection -- A New Baseline}, 
        year={2018} 
}