Research map / Video / Video segmentation and tracking
End-to-End Compressed Video Representation Learning for Generic Event Boundary Detection
Congcong Li, Xinyao Wang, Longyin Wen, Dexiang Hong, Tiejian Luo, Libo Zhang
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 the end-to-end frame feature strategy rather than repeated PC context-window inputs (ablation) | PC · Kinetics-GEBD local minval | F1: 0.621 -> 0.634 | Improved |
| Add Gaussian-smoothed soft labels (ablation) | PC + E2E · Kinetics-GEBD local minval | F1: 0.634 -> 0.654 | Improved |
| Add optical flow (ablation) | PC · Kinetics-GEBD local minval | F1: 0.621 -> 0.646 | Improved |
| Use SCCE to fuse I-frame features with compressed P-frame information (ablation) | PC · Kinetics-GEBD local minval | F1: 0.621 -> 0.669 | Improved |
| Add optical flow (ablation) | Ours · Kinetics-GEBD local minval | F1: 0.654 -> 0.661 | Improved |
| Use SCCE rather than a vanilla compressed-feature encoder (ablation) | Ours · Kinetics-GEBD local minval | F1: 0.657 -> 0.686 | Improved |
| Add the temporal contrastive module with window size k=2 (ablation) | Ours · Kinetics-GEBD local minval | F1: 0.686 -> 0.710 | Improved |
| Set temporal contrastive window size to k=8 instead of k=2 (ablation) | Ours · Kinetics-GEBD local minval | F1: 0.710 -> 0.746 | Improved |
| Increase the temporal contrastive window beyond k=8 (ablation) | Ours · Kinetics-GEBD local minval | recall: 0.757 (k=8) -> 0.725 (k=10); 0.696 (k=12) | Mixed |
| Use the proposed compressed-domain representation | Ours versus PC + Optical Flow · Kinetics-GEBD val, avg Rel.Dis. | F1: 0.830 -> 0.865 | Improved |
| Use SCCE-based compressed features instead of vanilla compressed features | Ours · UCF-101 | accuracy | Improved |
Closest papers by meaning
- Local Compressed Video Stream Learning for Generic Event Boundary Detection (IJCV 2023 · similarity 0.97)
- Generic Event Boundary Detection: A Benchmark for Event Segmentation (ICCV 2021 · similarity 0.88)
- Fine-grained Dynamic Network for Generic Event Boundary Detection (ECCV 2024 · similarity 0.88)
- UBoCo: Unsupervised Boundary Contrastive Learning for Generic Event Boundary Detection (CVPR 2022 · similarity 0.88)
- Temporal Action Detection Model Compression by Progressive Block Drop (CVPR 2025 · similarity 0.85)
- DiffDVC: Accurate Event Detection for Dense Video Captioning via Diffusion Models (AAAI 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.
