Weakly-Supervised Spatio-Temporal Anomaly Detection in Surveillance Video

Authors: Jie Wu, Wei Zhang, Guanbin Li, Wenhao Wu, Xiao Tan, Yingying Li, Errui Ding, Liang Lin

IJCAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct extensive qualitative and quantitative evaluations to demonstrate the effectiveness of the proposed approach and analyze the key factors that contribute more to handle this task.
Researcher Affiliation Collaboration 1Sun Yat-sen University 2Baidu Inc. 3Byte Dance Inc.
Pseudocode No The paper does not contain any pseudocode or clearly labeled algorithm blocks.
Open Source Code No The paper does not provide any explicit statements or links about the availability of source code.
Open Datasets No To this end, we build a new dataset (denoted as ST-UCF-Crime) that annotates spatio-temporal bounding boxes for abnormal events in UCF-Crime [Sultani et al., 2018]... Furthermore, we contribute a new dataset, namely Spatio-Temporal Road Accident (abbreviated as STRA)... The paper states they 'contribute' these datasets but does not provide concrete access information (e.g., a link, DOI, or repository) for the new annotations or the STRA dataset. It mentions 'We provide more details in the appendix', but the appendix is not provided.
Dataset Splits No The paper mentions 'We randomly choose 30 positive and 30 negative bags to construct a mini-batch' for training, and uses 'abnormal testing videos' for evaluation, but does not specify the explicit train/validation/test splits (e.g., percentages, sample counts, or references to predefined splits).
Hardware Specification No The paper does not explicitly describe the specific hardware used, such as GPU models, CPU models, or cloud computing instance types.
Software Dependencies No The paper mentions using 'C3D [Tran et al., 2015]' and 'Adam optimizer' but does not provide specific version numbers for these or other software libraries (e.g., Python, PyTorch, TensorFlow).
Experiment Setup Yes We randomly choose 30 positive and 30 negative bags to construct a mini-batch, and the number of instances per bag is limited to 200. The total loss is optimized via Adam optimizer with the learning rate of 0.0005.