Research map / Video / Video-language understanding
Multi-Stage Aggregated Transformer Network for Temporal Language Localization in Videos
Mingxing Zhang, Yang Yang, Xinghan Chen, Yanli Ji, Xing Xu, Jingjing Li, Heng Tao Shen
Check 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 |
|---|---|---|---|
| Use a visual-language transformer rather than Conv-LSTM with temporal convolution (ablation) | Temporal language localization model · ActivityNet Captions | Rank 1 @ IoU=0.5: 41.12 -> 46.96 | Improved |
| Add multi-stage aggregation rather than convolutional candidate features (ablation) | Conv-LSTM model · ActivityNet Captions | Rank 1 @ IoU=0.5: 41.12 -> 45.16 | Improved |
| Add multi-stage aggregation rather than mean pooling (ablation) | Visual-language transformer · ActivityNet Captions | Rank 1 @ IoU=0.5: 45.92 -> 46.96 | Improved |
| Add multi-stage aggregation rather than stacked convolution (ablation) | Visual-language transformer · ActivityNet Captions | Rank 1 @ IoU=0.5: 45.36 -> 46.96 | Improved |
| Add multi-stage aggregation rather than RoI pooling (ablation) | Visual-language transformer · ActivityNet Captions | Rank 1 @ IoU=0.5: 45.02 -> 46.96 | Improved |
| Decouple transformer parameters for visual and language modalities (ablation) | Visual-language transformer with MSA · ActivityNet Captions | Rank 1 @ IoU=0.5: 46.96 -> 48.02 | Improved |
| Use masked word prediction during training (ablation) | De-VLTrans-MSA · ActivityNet Captions | Rank 1 @ IoU=0.5: 46.26 -> 48.02 | Improved |
| Use three stages instead of only starting and ending stages (ablation) | De-VLTrans-MSA · ActivityNet Captions | Rank 1 @ IoU=0.5: 46.84 -> 48.02 | Improved |
| Use three stages rather than four moment stages (ablation) | De-VLTrans-MSA · ActivityNet Captions | Rank 1 @ IoU=0.5: 47.69 -> 48.02 | Improved |
| Increase shared-transformer parameters 4 times in width (ablation) | VLTrans-MSA · ActivityNet Captions | Rank 1 @ IoU=0.5: 46.96 -> 45.52 | Hurt |
| Increase shared-transformer parameters 4 times in depth (ablation) | VLTrans-MSA · ActivityNet Captions | Rank 1 @ IoU=0.5: 46.96 -> 20.74 | Hurt |
| Use two BERT streams instead of a decoupled single transformer (ablation) | De-VLTrans-MSA · ActivityNet Captions | Rank 1 @ IoU=0.5: 46.74 -> 48.02 | Hurt |
Closest papers by meaning
- Transferable Video Moment Localization by Moment-Guided Query Prompting (AAAI 2024 · similarity 0.90)
- Structured Multi-Level Interaction Network for Video Moment Localization via Language Query (CVPR 2021 · similarity 0.90)
- Unsupervised Domain Adaptative Temporal Sentence Localization with Mutual Information Maximization (AAAI 2024 · similarity 0.87)
- Boundary-Aware Temporal Dynamic Pseudo-Supervision Pairs Generation for Zero-Shot Natural Language Video Localization (AAAI 2025 · similarity 0.87)
- Structured Video-Language Modeling with Temporal Grouping and Spatial Grounding (ICLR 2024 · similarity 0.87)
- Universal Video Temporal Grounding with Generative Multi-modal Large Language Models (NeurIPS 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.
