Related Experiment Videos
Mamba-based modulated fusion model for video moment retrieval
Bing Yu1,2, Jingyu Li3,4, Youxian Di3,4
1Department of Film and Television Engineering, Shanghai University, Shanghai, 200072, China. yubing@shu.edu.cn.
Scientific Reports
|April 3, 2026
Summary
This study introduces the Hybrid Mamba Network (HM-Net) for improved video moment retrieval (VMR). HM-Net enhances long-range temporal reasoning, significantly boosting accuracy in localizing moments within videos.
Area of Science:
- Computer Science
- Artificial Intelligence
- Machine Learning
Background:
- Video Moment Retrieval (VMR) is crucial for video understanding, connecting text queries to video segments.
- Current VMR methods struggle with global temporal context, especially in long videos.
Purpose of the Study:
- To develop an advanced architecture for VMR that effectively captures long-range temporal dependencies.
- To improve the accuracy of localizing relevant video moments from textual descriptions.
Main Methods:
- Proposed Hybrid Mamba Network (HM-Net), a two-level fusion architecture.
- Introduced the Hybrid Modulated Bi-Mamba (HMB) Block, integrating Mamba for enhanced temporal reasoning.
- Utilized attention and sequence modeling for comprehensive video analysis.
Main Results:
- HM-Net achieved superior performance on TACoS and QVHighlights benchmarks.
- Demonstrated a 3.84% improvement in R1@0.5 on TACoS and 1.65% in mAP on QVHighlights.
- Showcased significant gains in localization accuracy, particularly for long-form videos.
Conclusions:
- HM-Net effectively addresses limitations in capturing global temporal context for VMR.
- The proposed architecture offers a promising direction for advancing video understanding and retrieval.
- HM-Net provides state-of-the-art performance in video moment localization.
Related Concept Videos
01:34The Fluid Mosaic Model
185.3K
The fluid mosaic model was first proposed as a visual representation of research observations. The model comprises the composition and dynamics of membranes and serves as a foundation for future membrane-related studies. The model depicts the structure of the plasma membrane with a variety of components, which include phospholipids, proteins, and carbohydrates. These integral molecules are loosely bound, defining the cell’s border and providing fluidity for optimal function.
185.3K
01:09Moment-of-Momentum Equation
536
The moment-of-momentum equation is a critical tool for analyzing the torque produced by the rotating blades of a wind turbine. This equation is derived by applying Newton's second law to a fluid particle, which states that the rate of change of linear momentum is equal to the external force acting on the particle.
536