Aug 2026· Conference on Multimedia Information Processing and Retrieval· pp. 225-229· 0 citations· 15 references
Computer ScienceEngineering
TL;DR
A lightweight two-stage framework for real-time video anomaly detection that eliminates the need for optical flow, standalone pose estimators, and density-based scoring modules is proposed.
Abstract
We propose a lightweight two-stage framework for real-time video anomaly detection. The first stage employs YOLO v11n-pose to detect persons and extract seventeen skeletal keypoints in a single forward pass. The second stage encodes each cropped person region through CLIP ViT-B/32 and computes cosine similarity against predefined textual descriptions of anomalous behaviors. This architecture eliminates the need for optical flow, standalone pose estimators, and density-based scoring modules. Experiments on CUHK Avenue, ShanghaiTech Campus, and a custom indoor dataset collected at Chulalongkorn University demonstrate an end-to-end throughput of approximately 51 FPS on an NVIDIA Titan XP GPU, a 3.36× speedup over the multi-feature baseline, while maintaining frame-level AUROC values of 89.26%, 70.26%, and 84.13%, respectively.
STEP, a simple framework that utilizes Principal Component Analysis (PCA) to project pose sequences into a compact, whitened PC-space, outperforms the previous skeleton-based state-of-the-art by 12.2% on the challenging UBnormal dataset and achieves highly competitive results by improving on the ShanghaiTech benchmark.
Jakub Micorek, M. Koziński, Horst Possegger· 0 citations
Experimental results show that the evaluation quantifies both benchmark anomaly detection and direct known-versus-unknown separation under the declared protocols, and that object prompts dominate the fused decision, while the auxiliary prompt groups do not yet contribute positively under the held-out-category protocol.
You-Xi Li, Xiang Chen, Li-Min Wang et al.· Scientific Reports· 0 citations
This research introduces a novel memory-augmented CNN-ConvViT autoencoder framework for unsupervised video anomaly detection and introduces a Temporal-Aware Prototype Memory Module (TAPMM) that explicitly learns normal spatio-temporal behavior patterns.
FASTe introduces a LogSumExp-based multiple instance learning (MIL) aggregation strategy for robust training on variable-length inputs, and introduces frame-level anomaly localization under weak supervision, without requiring dense labels.
Jihun Jeon, Raeyoung Chang, Jisu Kim et al.· IEEE Access· 0 citations
Automated anomaly detection in video streams is essential for industrial process monitoring, security surveillance, and quality control systems. This paper presents a systematic comparative analysis of two state-of-the-art deep learning object detectors, YOLOv8 and Faster R-CNN, applied to the task of visual anomaly de...
S. Pchelintseva, S. Shiryaeva, I. Vakhrusheva· 2026 International Russian A...· 0 citations
To address the trade-off between detection accuracy and deployment efficiency in surveillance-based anomaly detection, we propose a lightweight multimodal abnormal behavior detection method based on GhostNet and FiLM fusion (G-FLM). GhostNet is employed as a lightweight visual backbone to reduce computational redundanc...
Bo Yin, Ming-Quan Zhang· International Journal of Adv...· 0 citations
Known for his clear and elegant writing style, Bertsekas shaped fields from control and optimization to large-scale computation and artificial intelligence.
