Research map / Video / Video-language understanding
Weakly Supervised Temporal Sentence Grounding With Gaussian-Based Contrastive Proposal Learning
Check 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 |
|---|---|---|---|
| Use learnable Gaussian proposals and intra-video contrastive proposal learning | CPL · Charades-STA test | R@1, IoU=0.5: 39.19 -> 49.24 | Helped |
| Use CPL with vote-based inference and α2=0.1 | CPL · ActivityNet Captions val 2 | R@1, IoU=0.1: 81.61 -> 82.55 | Helped |
| Add the intra-video contrastive loss and diversity loss (ablation) | CPL · Charades-STA test | R@1, IoU=0.5: 20.49 -> 49.24 | Helped |
| Add the diversity loss Ldiv (ablation) | CPL · Charades-STA test | R@1, IoU=0.5: 32.08 -> 49.24 | Helped |
| Use learnable positive proposals instead of fixed proposals (ablation) | CPL · Charades-STA test | R@1, IoU=0.5: 28.97 -> 45.47 | Helped |
| Add negative proposals to learnable positive proposals (ablation) | CPL · Charades-STA test | R@1, IoU=0.5: 45.47 -> 49.24 | Helped |
| Use intra-video negatives instead of negatives from other videos (ablation) | CPL · Charades-STA test | R@1, IoU=0.5: 47.56 -> 49.24 | Helped |
| Optimize only the positive proposal with the smallest reconstruction loss (ablation) | CPL · Charades-STA test | R@1, IoU=0.5: 36.16 -> 49.24 | Helped |
| Set α2=1 rather than α2=0.1 (ablation) | CPL · ActivityNet Captions val 2 | R@1, IoU=0.1: 82.55 -> 71.23 | Hurt |
| Set α2=1 rather than α2=0.1 (ablation) | CPL · ActivityNet Captions val 2 | R@5, IoU=0.5: 43.13 -> 65.79 | Helped |
| Mine negative proposals dynamically from easy to hard rather than keeping them far away or always close (ablation) | CPL · Charades-STA | mIoU | Helped |
Closest papers by meaning
- Universal Video Temporal Grounding with Generative Multi-modal Large Language Models (NeurIPS 2025 · similarity 0.87)
- Siamese Learning with Joint Alignment and Regression for Weakly-Supervised Video Paragraph Grounding (CVPR 2024 · similarity 0.86)
- Learning Multi-Scale Video-Text Correspondence for Weakly Supervised Temporal Article Gronding (AAAI 2024 · similarity 0.86)
- Gaussian Mixture Proposals with Pull-Push Learning Scheme to Capture Diverse Events for Weakly Supervised Temporal Video Grounding (AAAI 2024 · similarity 0.86)
- Multi-Scale Contrastive Learning for Video Temporal Grounding (AAAI 2025 · similarity 0.86)
- Multi-Pair Temporal Sentence Grounding via Multi-Thread Knowledge Transfer Network (AAAI 2025 · similarity 0.86)
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.