Research map / Video / Video-language understanding
Unsupervised Temporal Video Grounding with Deep Semantic Clustering
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Add neck-diversity regularization L_dqa (ablation) | DSCNet language and video modules · Charades-STA | R@1 IoU=0.3: 28.38 -> 33.54 | Helped |
| Add sentence-level semantic loss L_mse (ablation) | DSCNet language module · Charades-STA | R@1 IoU=0.3: 33.54 -> 36.78 | Helped |
| Add foreground triplet loss L_trip (ablation) | DSCNet video module · Charades-STA | R@1 IoU=0.3: 36.78 -> 40.23 | Helped |
| Add specific-attention loss L_sab (ablation) | DSCNet video module · Charades-STA | R@1 IoU=0.3: 36.78 -> 41.97 | Helped |
| Combine L_sab and L_trip in the video module (ablation) | DSCNet · Charades-STA | R@1 IoU=0.3: 41.97 -> 44.15 | Helped |
| Use averaged cluster centers rather than random center selection (ablation) | DSCNet semantic-center selection · Charades-STA | R@1 IoU=0.3: 23.79 -> 44.15 | Helped |
| Use averaged cluster centers rather than a random sample from each cluster (ablation) | DSCNet semantic-center selection · Charades-STA | R@1 IoU=0.3: 39.67 -> 44.15 | Helped |
| Use 4 semantic necks rather than 2 (ablation) | DSCNet language module · Charades-STA | R@1 IoU=0.3: 33.06 -> 44.15 | Helped |
| Use 4 semantic necks rather than 8 (ablation) | DSCNet language module · Charades-STA | R@1 IoU=0.3: 40.90 -> 44.15 | Helped |
| Increase iterative-learning training from 1 to 5 iterations (ablation) | DSCNet video module · Charades-STA | R@1 IoU=0.3: 29.38 -> 44.15 | Helped |
| Increase iterative-learning training from 5 to 7 iterations (ablation) | DSCNet video module · Charades-STA | R@1 IoU=0.3: 44.15 -> 44.06 | No clear effect |
| Use DSCNet without paired video-query annotations | DSCNet · ActivityNet Captions | R@1 IoU=0.3: DSCNet 47.29; VGN 50.12 | Mixed |
Closest papers by meaning
- VideoGrounding-DINO: Towards Open-Vocabulary Spatio-Temporal Video Grounding (CVPR 2024 · similarity 0.89)
- Empower Words: DualGround for Structured Phrase and Sentence-Level Temporal Grounding (NeurIPS 2025 · similarity 0.89)
- VideoGEM: Training-free Action Grounding in Videos (CVPR 2025 · similarity 0.89)
- Universal Video Temporal Grounding with Generative Multi-modal Large Language Models (NeurIPS 2025 · similarity 0.89)
- Comprehensive Visual Grounding for Video Description (AAAI 2024 · similarity 0.89)
- Multi-Pair Temporal Sentence Grounding via Multi-Thread Knowledge Transfer Network (AAAI 2025 · similarity 0.89)
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.