Research map / Video / Video-language understanding
BAM-DETR: Boundary-Aligned Moment Detection Transformer for Temporal Sentence Grounding in Videos
Pilhyeon Lee, Hyeran Byun
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 |
|---|---|---|---|
| replace center-width moment formulation with boundary-oriented (anchor, d_s, d_e) modeling (ablation) | BAM-DETR baseline · QVHighlights val | mAP avg: 41.75 -> 42.42 | Improved |
| add dual-pathway decoder (ablation) | BAM-DETR · QVHighlights val | mAP avg: 42.42 -> 44.16 | Improved |
| add localization quality-based scoring instead of classification scores (ablation) | BAM-DETR · QVHighlights val | mAP avg: 44.16 -> 47.61 | Improved |
| sample boundary features dynamically with offsets instead of fixed boundary points (ablation) | boundary-focused attention · QVHighlights val | mAP avg: 45.17 -> 47.14 | Improved |
| increase boundary points from 3 to 5 (ablation) | boundary-focused attention · QVHighlights val | mAP avg: 47.61 -> 46.70 | No clear effect |
| use focused attention for boundary queries instead of global attention (ablation) | BAM-DETR decoder · QVHighlights val | mAP avg: 44.54 -> 46.21 | Improved |
| use focused attention for anchor queries too (ablation) | BAM-DETR decoder · QVHighlights val | mAP avg: 46.21 -> 45.87 | Hurt |
| use anchor, start and end queries together for quality prediction (ablation) | BAM-DETR · QVHighlights val | mAP avg: 46.26 -> 47.61 | Improved |
| use merged memory features (plain plus boundary-sensitive) for focused attention (ablation) | BAM-DETR · QVHighlights val | mAP avg: 46.21 -> 47.61 | Improved |
| increase number of predictions M from 5 to 10 and above (ablation) | BAM-DETR · QVHighlights val | mAP avg: 45.18 -> 47.61 | Improved |
| BAM-DETR vs previous best on anti-biased Charades-STA length split | BAM-DETR · Anti-biased Charades-STA (moment length) | R1@0.7: 32.87 -> 40.74 | Improved |
Later papers that compared with this method
6 later accepted papers report BAM-DETR as a baseline in their results tables; 6 of them report a better number on most of the results they share with it.
- Decoupling Moment from Event for Video Temporal Grounding (ECCV 2026 · better on 4 of 4 shared results)
- CVA: Context-aware Video-text Alignment for Video Temporal Grounding (CVPR 2026 · better on 3 of 3 shared results)
- Object-Centric Framework for Video Moment Retrieval (AAAI 2026 · better on 2 of 2 shared results)
- DE2TR: Dual Evidence Detection Transformer for Video Temporal Grounding (ECCV 2026 · better on 1 of 1 shared result)
- The Devil is in the Spurious Correlations: Boosting Moment Retrieval with Dynamic Learning (ICCV 2025 · better on 6 of 6 shared results)
- Sim-DETR: Unlock DETR for Temporal Sentence Grounding (ICCV 2025 · better on 5 of 5 shared results)
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
- Sim-DETR: Unlock DETR for Temporal Sentence Grounding (ICCV 2025 · similarity 0.91)
- Transferable Video Moment Localization by Moment-Guided Query Prompting (AAAI 2024 · similarity 0.87)
- Empower Words: DualGround for Structured Phrase and Sentence-Level Temporal Grounding (NeurIPS 2025 · similarity 0.87)
- Diversifying Query: Region-Guided Transformer for Temporal Sentence Grounding (AAAI 2025 · similarity 0.87)
- Zero-shot Video Moment Retrieval via Off-the-shelf Multimodal Large Language Models (AAAI 2025 · similarity 0.86)
- Universal Video Temporal Grounding with Generative Multi-modal Large Language Models (NeurIPS 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.
