Research map / Video / Video-language understanding
ProTeGe: Untrimmed Pretraining for Video Temporal Grounding by Video Temporal Grounding
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 |
|---|---|---|---|
| Pretrain on untrimmed videos using VTG rather than trimmed video-text alignment | ProTéGé features with 2D-TAN · Charades-STA | R@0.5: 46.11 -> 53.26 | Helped |
| Pretrain on untrimmed videos using VTG rather than trimmed video-text alignment | ProTéGé features with 2D-TAN · ActivityNet-Captions | R@0.7: 26.51 -> 29.02 | Helped |
| Use ProTéGé-pretrained features | CPL on Charades-STA · Charades-STA weakly supervised | mIoU: 42.78 -> 43.76 | Helped |
| Use ProTéGé-pretrained features | CNM on ActivityNet-Captions · ActivityNet-Captions weakly supervised | mIoU: +3.15% | Helped |
| Use ProTéGé-pretrained features | PSVL · Charades-STA zero-shot | mIoU: 30.09 -> 31.25 | Helped |
| Use ProTéGé-pretrained features | PSVL · ActivityNet-Captions zero-shot | mIoU: 30.78 -> 33.04 | Helped |
| Add localization loss to the pretraining objective (ablation) | ProTéGé · Charades-STA fully supervised | R@0.5: 49.40 -> 53.26 | Helped |
| Add inter- and intra-video-text alignment losses (ablation) | ProTéGé · Charades-STA fully supervised | R@0.5: 51.63 -> 53.26 | Helped |
| Use VT-SGM proposal grid instead of direct timestamp regression (ablation) | ProTéGé · Charades-STA fully supervised | R@0.5: 48.36 -> 53.26 | Helped |
| Pretrain on untrimmed videos via VTG rather than directly aligning video and text (ablation) | ProTéGé · Charades-STA fully supervised | R@0.5: 47.35 -> 53.26 | Helped |
| Add cross-entropy localization loss (ablation) | ProTéGé · Charades-STA fully supervised | R@0.5: 51.18 -> 53.26 | Helped |
| Add distance-based localization loss (ablation) | ProTéGé · Charades-STA fully supervised | R@0.5: 51.86 -> 53.26 | Helped |
Closest papers by meaning
- Empower Words: DualGround for Structured Phrase and Sentence-Level Temporal Grounding (NeurIPS 2025 · similarity 0.91)
- VTG-LLM: Integrating Timestamp Knowledge into Video LLMs for Enhanced Video Temporal Grounding (AAAI 2025 · similarity 0.91)
- Siamese Learning with Joint Alignment and Regression for Weakly-Supervised Video Paragraph Grounding (CVPR 2024 · similarity 0.90)
- Learning Multi-Scale Video-Text Correspondence for Weakly Supervised Temporal Article Gronding (AAAI 2024 · similarity 0.89)
- TRACE: Temporal Grounding Video LLM via Causal Event Modeling (ICLR 2025 · similarity 0.88)
- ReVisionLLM: Recursive Vision-Language Model for Temporal Grounding in Hour-Long Videos (CVPR 2025 · similarity 0.88)
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.