Research map / Video / Video-language understanding
UBoCo: Unsupervised Boundary Contrastive Learning for Generic Event Boundary Detection
Hyolim Kang, Jinwoo Kim, Taehyun Kim, Seon Joo Kim
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 UBoCo instead of the previous unsupervised PA method | Unsupervised GEBD · Kinetics-GEBD validation | F1@0.05: 39.6 -> 70.3 | Improved |
| Use UBoCo instead of the previous supervised PC baseline | GEBD · Kinetics-GEBD validation | F1@0.05: 62.5 -> 70.3 | Improved |
| Replace RTP with simple thresholding (ablation) | UBoCo · Kinetics-GEBD validation | F1@0.05: 70.3 -> 27.3; 70.2 -> 27.3 | Hurt |
| Replace RTP with local maxima (ablation) | UBoCo · Kinetics-GEBD validation | F1@0.05: 70.3 -> 68.1; 70.2 -> 68.8 | Hurt |
| Add zero-padding to RTP (ablation) | UBoCo · Kinetics-GEBD validation | F1@0.05: 55.3 -> 70.3; 54.8 -> 70.2 | Improved |
| Add ground-truth supervision (ablation) | UBoCo · Kinetics-GEBD validation | F1@0.05: 70.3 -> 71.1 (+0.8); 70.2 -> 75.5 (+5.3) | Improved |
| Add a decoder to the supervised model (ablation) | SBoCo · Kinetics-GEBD validation | F1@0.05: 71.1 -> 73.2 (+2.9); 75.5 -> 78.7 (+8.5) | Improved |
| Add BoCo loss alongside BCE loss (ablation) | SBoCo · Kinetics-GEBD validation | F1@0.05: 71.8 -> 73.2 (+1.4); 77.5 -> 78.7 (+1.2) | Improved |
| Train the encoder with self-supervision using pseudo-labels and BoCo loss (ablation) | UBoCo-Res50 · Kinetics-GEBD validation | F1@0.05: 33.2 -> 70.3 | Improved |
| Use TSN video-level features instead of ResNet-50 image-level features (ablation) | SBoCo · Kinetics-GEBD validation | F1@0.05: 73.2 -> 78.7 | Improved |
Closest papers by meaning
- Fine-grained Dynamic Network for Generic Event Boundary Detection (ECCV 2024 · similarity 0.90)
- Online Generic Event Boundary Detection (ICCV 2025 · similarity 0.89)
- Generic Event Boundary Detection via Denoising Diffusion (ICCV 2025 · similarity 0.89)
- DIBS: Enhancing Dense Video Captioning with Unlabeled Videos via Pseudo Boundary Enrichment and Online Refinement (CVPR 2024 · similarity 0.84)
- LongVALE: Vision-Audio-Language-Event Benchmark Towards Time-Aware Omni-Modal Perception of Long Videos (CVPR 2025 · similarity 0.83)
- VADTree: Explainable Training-Free Video Anomaly Detection via Hierarchical Granularity-Aware Tree (NeurIPS 2025 · similarity 0.83)
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.
