Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
October 16, 2025Open Access

Mixture of Experts Guided by Gaussian Splatters Matters: A new Approach to Weakly-Supervised Video Anomaly Detection

View Full Paper
Ask AI
Bookmark
Share

Authors

GDGiacomo D’AmicantonioSMSnehashis MajhiQKQuan Kong

Discussion

Loading...

Member takes

Overview

New framework improves anomaly detection in videos by leveraging specialized models and temporal guidance.

Key Points

  • The GS-MoE framework achieves 91.58% AUC on the UCF-Crime dataset, demonstrating significant performance gains.
  • By integrating a set of expert models, the approach captures diverse anomaly types more effectively than existing methods.
  • Temporal guidance through Gaussian splatting enhances the model’s ability to identify nuanced patterns of anomalies.
  • Integration of specialized predictions allows for capturing complex relationships in various video anomalies.

Cite This Study

D’Amicantonio et al. (2025) studied this question.

synapsesocial.com/papers/68f12bfb2107091eab27a47ehttps://doi.org/10.48550/arxiv.2508.06318
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1TASG-VAD: Weakly Supervised Video Anomaly Detection via Temporal Variation Attention and Adaptive Saliency Guidance2026
  2. 2Learn Suspected Anomalies from Event Prompts for Video Anomaly Detection2024
  3. 3Distilling Aggregated Knowledge for Weakly-Supervised Video Anomaly Detection2024
  4. 4MMVAD: A Vision-Language Model for Cross-Domain Video Anomaly Detection with Contrastive Learning and Scale-Adaptive Frame Segmentation2024 · 1 citations
  5. 5Temporal-Enhanced and Visual-Text Adaptive Fusion for Weakly Supervised Video Anomaly Detection in Public Safety2026