Research map / Video / Video-language understanding
DE2TR: Dual Evidence Detection Transformer for Video Temporal Grounding
Yifan Zhang, Chengxu Liu, Yujie Dun, Xueming Qian
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 DEFormer dual-branch decoder only (ablation) | DE2TR baseline · QVHighlights val | MR mAP Avg.: +0.67% | Improved |
| add semantic and boundary evidence alignment losses to DEFormer (ablation) | DE2TR · QVHighlights val | MR mAP Avg.: +3.61% absolute over the baseline | Improved |
| add prior-guided refinement mechanism (ablation) | DE2TR · QVHighlights val | MR mAP Avg.: +1.43% | Improved |
| add boundary augmentation strategy (ablation) | DE2TR · QVHighlights val | MR mAP Avg.: 50.67 -> 52.12 | Improved |
| all components together (ablation) | DE2TR · QVHighlights val | MR mAP Avg.: +6.49% overall | Improved |
| route decoupled queries separately to heads instead of a merged query (ablation) | DE2TR · QVHighlights val | MR mAP Avg.: 49.53 -> 50.67 | Improved |
| inject semantic prior into queries and boundary prior into memory (vs both memory) (ablation) | DE2TR · QVHighlights val | MR mAP Avg.: 50.14 -> 50.67 | Improved |
| inject semantic prior into memory and boundary into queries (swapped) (ablation) | DE2TR · QVHighlights val | MR mAP Avg.: 50.67 -> 49.86 | Hurt |
| increase moment queries from 10 to 15 (ablation) | DE2TR · QVHighlights val | MR mAP Avg.: 50.67 -> 51.22 | Improved |
| increase moment queries from 15 to 20 (ablation) | DE2TR · QVHighlights val | MR mAP Avg.: 51.22 -> 51.19 | No clear effect |
| reduce moment queries from 10 to 5 (ablation) | DE2TR · QVHighlights val | MR mAP Avg.: 50.67 -> 48.56 | Hurt |
| use InternVideo2 features instead of SlowFast+CLIP (ablation) | DE2TR · QVHighlights val | MR mAP Avg.: 50.67 -> 53.91 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- BAM-DETR — BAM-DETR: Boundary-Aligned Moment Detection Transformer for Temporal Sentence Grounding in Videos (ECCV 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Keeping the Evidence Chain: Semantic Evidence Allocation for Training-Free Token Pruning in Video Temporal Grounding (ECCV 2026 · similarity 0.91)
- Foresee-to-Ground: From Predictive Temporal Perception to Evidence-Driven Reasoning for Video Temporal Grounding (ICML 2026 · similarity 0.91)
- SE-DETR: Explicit Semantic Exploration for Generalizability and Distinguishability in Video Temporal Grounding (ECCV 2026 · similarity 0.91)
- CausalVTG: Towards Robust Video Temporal Grounding via Causal Inference (NeurIPS 2025 · similarity 0.90)
- Empower Words: DualGround for Structured Phrase and Sentence-Level Temporal Grounding (NeurIPS 2025 · similarity 0.89)
- Boundary-Aware Temporal Dynamic Pseudo-Supervision Pairs Generation for Zero-Shot Natural Language Video Localization (AAAI 2025 · similarity 0.87)
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.
