Research map / Video / Video segmentation and tracking
Progressive Attention on Multi-Level Dense Difference Maps for Generic Event Boundary Detection
Jiaqi Tang, Zhaoyang Liu, Chen Qian, Wayne Wu, Limin Wang
Check your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use dense difference maps with RGB features instead of RGB features alone (ablation) | DDM-Net · Kinetics-GEBD | F1@0.05: 0.6793 -> 0.7643 | Improved |
| Replace optical flow with DDM in the motion representation (ablation) | DDM-Net · Kinetics-GEBD | F1@0.05: 0.6625 -> 0.7512 | Improved |
| Combine RGB with DDM instead of using a two-stream RGB and optical-flow representation (ablation) | DDM-Net · Kinetics-GEBD | F1@0.05: 0.6881 -> 0.7643 | Improved |
| Use a local clip with stride s=6 rather than a much wider view with s=30 (ablation) | DDM-Net · Kinetics-GEBD | F1@0.05: 0.6914 -> 0.7643 | Improved |
| Use more frames, with w=15 instead of w=5 (ablation) | DDM-Net · Kinetics-GEBD | F1@0.05: 0.7643 -> 0.7703 | Improved |
| Add multi-level spatial and temporal features instead of using a single feature level (ablation) | DDM-Net · Kinetics-GEBD | F1@0.05: 0.7353 -> 0.7643 | Improved |
| Replace Euclidean distance with Manhattan distance for DDM (ablation) | DDM-Net · Kinetics-GEBD | F1@0.05: 0.7643 -> 0.7632 | No clear effect |
| Replace Euclidean distance with Chebyshev distance for DDM (ablation) | DDM-Net · Kinetics-GEBD | F1@0.05: 0.7643 -> 0.7540 | Hurt |
| Add both intra-modal and cross-modal attention instead of averaging features (ablation) | DDM-Net · Kinetics-GEBD | F1@0.05: 0.7498 -> 0.7643 | Improved |
| Replace pyramidal feature hierarchy with FPN (ablation) | DDM-Net · Kinetics-GEBD | F1@0.05: 0.7643 -> 0.7511 | Hurt |
| Increase intra-modal and cross-modal attention layers from 1 each to 6 each (ablation) | DDM-Net · Kinetics-GEBD | F1@0.05: 0.7584 -> 0.7643 | Improved |
| Use 5 learnable queries rather than 1 or 10 (ablation) | DDM-Net · Kinetics-GEBD | F1@0.05: 0.7579 -> 0.7643; 0.7592 -> 0.7643 | Improved |
Closest papers by meaning
- DiffDVC: Accurate Event Detection for Dense Video Captioning via Diffusion Models (AAAI 2025 · similarity 0.86)
- Dense Audio-Visual Event Localization Under Cross-Modal Consistency and Multi-Temporal Granularity Collaboration (AAAI 2025 · similarity 0.86)
- DELTAVID: Enhancing Fine-Grained Spatiotemporal Perception with Cross-Video Differences (NeurIPS 2026 · similarity 0.86)
- Adapting to the Unknown: Training-Free Audio-Visual Event Perception with Dynamic Thresholds (CVPR 2025 · similarity 0.85)
- MotionBench: Benchmarking and Improving Fine-grained Video Motion Understanding for Vision Language Models (CVPR 2025 · similarity 0.85)
- Event-Equalized Dense Video Captioning (CVPR 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
