@article{ART003246052},
author={Da-Jeong Seo and Yoo-Sung Kim},
title={Analysis of Strategies for Using Limited Data to Recognize Abnormal Movements in Crowds},
journal={Journal of The Korea Society of Computer and Information},
issn={1598-849X},
year={2025},
volume={30},
number={9},
pages={53-62}
TY - JOUR
AU - Da-Jeong Seo
AU - Yoo-Sung Kim
TI - Analysis of Strategies for Using Limited Data to Recognize Abnormal Movements in Crowds
JO - Journal of The Korea Society of Computer and Information
PY - 2025
VL - 30
IS - 9
PB - The Korean Society Of Computer And Information
SP - 53
EP - 62
SN - 1598-849X
AB - This study investigates strategies for effectively utilizing limited data to recognize abnormal movements in crowd scenarios, based on a newly constructed crowd movement video dataset. Existing datasets such as Crowd-11 dataset focus on general flow patterns and lacks suitability for detecting abnormal behaviors. To address this, we redefined classification criteria and expanded the dataset with additional video samples. We tested various data composition strategies by changing the number of sequences per video and data splitting methods, and evaluated their impacts on recognition performance.
Using optical flow for motion feature extraction and a Bidirectional LSTM model, we observed that more sequences led to improved performance, while fixed data splits showed high accuracy but risked overfitting. The proposed dataset and findings provide a solid foundation for developing abnormal crowd behavior detection models and offer practical guidance for data design in real-world applications.
KW - Abnormal Movement Recognition;Construction of Crowd Movement Dataset;;Definition of Movement Classes;Data Usage Strategy
DO -
UR -
ER -
Da-Jeong Seo and Yoo-Sung Kim. (2025). Analysis of Strategies for Using Limited Data to Recognize Abnormal Movements in Crowds. Journal of The Korea Society of Computer and Information, 30(9), 53-62.
Da-Jeong Seo and Yoo-Sung Kim. 2025, "Analysis of Strategies for Using Limited Data to Recognize Abnormal Movements in Crowds", Journal of The Korea Society of Computer and Information, vol.30, no.9 pp.53-62.
Da-Jeong Seo, Yoo-Sung Kim "Analysis of Strategies for Using Limited Data to Recognize Abnormal Movements in Crowds" Journal of The Korea Society of Computer and Information 30.9 pp.53-62 (2025) : 53.
Da-Jeong Seo, Yoo-Sung Kim. Analysis of Strategies for Using Limited Data to Recognize Abnormal Movements in Crowds. 2025; 30(9), 53-62.
Da-Jeong Seo and Yoo-Sung Kim. "Analysis of Strategies for Using Limited Data to Recognize Abnormal Movements in Crowds" Journal of The Korea Society of Computer and Information 30, no.9 (2025) : 53-62.
Da-Jeong Seo; Yoo-Sung Kim. Analysis of Strategies for Using Limited Data to Recognize Abnormal Movements in Crowds. Journal of The Korea Society of Computer and Information, 30(9), 53-62.
Da-Jeong Seo; Yoo-Sung Kim. Analysis of Strategies for Using Limited Data to Recognize Abnormal Movements in Crowds. Journal of The Korea Society of Computer and Information. 2025; 30(9) 53-62.
Da-Jeong Seo, Yoo-Sung Kim. Analysis of Strategies for Using Limited Data to Recognize Abnormal Movements in Crowds. 2025; 30(9), 53-62.
Da-Jeong Seo and Yoo-Sung Kim. "Analysis of Strategies for Using Limited Data to Recognize Abnormal Movements in Crowds" Journal of The Korea Society of Computer and Information 30, no.9 (2025) : 53-62.
