Research map / Video / Video-language understanding
E3M: Zero-Shot Spatio-Temporal Video Grounding with Expectation-Maximization Multimodal Modulation
Peijun Bao, Zihao Shao, Wenhan Yang, Boon Poh Ng, Alex Kot
Read the paperCheck 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 |
|---|---|---|---|
| add EM alternating optimization of temporal and spatial grounding (ablation) | E3M · HC-STVG | m_vIoU: 17.21 -> 19.11 | Improved |
| add context-based visual modulation (ablation) | E3M · HC-STVG | m_vIoU: 17.05 -> 19.11 | Improved |
| add prototype-based textual modulation (ablation) | E3M · HC-STVG | m_vIoU: 18.76 -> 19.11 | Improved |
| use long context (T=10) instead of short context (1) (ablation) | visual modulation · HC-STVG | m_vIoU: 17.62 -> 19.11 | Improved |
| use both past and future context instead of one side (ablation) | visual modulation · HC-STVG | m_vIoU: 18.27 (past only), 18.09 (future only) -> 19.11 | Improved |
| use K=3 prototypes instead of a single one (ablation) | textual modulation · HC-STVG | m_vIoU: 18.80 -> 19.11 | Improved |
| increase prototype count K beyond 3 (ablation) | textual modulation · HC-STVG | avg vIoU@{0.3,0.5}: 20.00 (K=3) -> 20.09 (K=5) | No clear effect |
| increase context size T beyond 10 (ablation) | visual modulation · HC-STVG | avg vIoU@{0.3,0.5}: 20.14 -> 20.22 | No clear effect |
| use CLIP RN-50 instead of ViT-B/32 (ablation) | E3M · HC-STVG | m_vIoU: 19.11 -> 18.29 | Hurt |
| zero-shot E3M versus weakly supervised Winner | HC-STVG · HC-STVG | vIoU@0.5: 6.12 -> 10.60 | Improved |
Closest papers by meaning
- Unleashing the Potential of Multimodal LLMs for Zero-Shot Spatio-Temporal Video Grounding (NeurIPS 2025 · similarity 0.93)
- VideoGrounding-DINO: Towards Open-Vocabulary Spatio-Temporal Video Grounding (CVPR 2024 · similarity 0.89)
- VideoGEM: Training-free Action Grounding in Videos (CVPR 2025 · similarity 0.89)
- STPro: Spatial and Temporal Progressive Learning for Weakly Supervised Spatio-Temporal Grounding (CVPR 2025 · similarity 0.88)
- Contextual Self-paced Learning for Weakly Supervised Spatio-Temporal Video Grounding (ICLR 2025 · similarity 0.88)
- What When and Where? Self-Supervised Spatio-Temporal Grounding in Untrimmed Multi-Action Videos from Narrated Instructions (CVPR 2024 · 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.
