Temporal Action Proposal Generation with Attention Mechanism
Author Information
Author(s): Sooksatra Sorn, Watcharapinchai Sitapa
Primary Institution: National Electronic and Computer Technology Center, National Science and Technology Development Agency, Thailand
Hypothesis
The TAPG model could achieve higher performance for small action proposals by emphasizing the regions of interest in the attention mask.
Conclusion
The proposed method significantly improves the performance of short-duration and contiguous action proposals.
Supporting Evidence
- The proposed G-MCBD model achieved an average recall of 78.22%.
- The method was particularly effective for action proposals with small temporal lengths.
- Empirical results showed improved performance in videos with small gap displacements.
Takeaway
This study helps computers better understand actions in long videos by focusing on small gaps between actions, making it easier to identify them.
Methodology
The study used a gap-aware attention mechanism to improve the accuracy of action proposal generation from untrimmed videos.
Limitations
The method may misclassify long action instances as short actions and struggles with overlapping proposals.
Participant Demographics
The dataset comprised 19,994 videos from YouTube, with an average video length of 115 seconds.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website