Research map / Video / Video-language understanding
Towards Bridging Event Captioner and Sentence Localizer for Weakly Supervised Dense Event Captioning
Shaoxiang Chen, Yu-Gang Jiang
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 |
|---|---|---|---|
| bridge the captioner and localizer with concept features, word-importance guidance, and proposal reconstruction | WS-DEC · ActivityNet Captions validation | CIDEr: 18.77 -> 21.21 | Improved |
| bridge the captioner and localizer with concept features, word-importance guidance, and proposal reconstruction | WS-DEC · ActivityNet Captions, top-1 recall at IoU=0.5 | Top-1 recall: 23.34 -> 24.16 | Improved |
| add MIL concept learning (ablation) | EC-SL base · ActivityNet Captions validation | CIDEr: 19.09 -> 20.46 | Improved |
| add ISAB to MIL concept learning (ablation) | EC-SL · ActivityNet Captions validation | CIDEr: 20.46 -> 20.98 | Improved |
| add gating to the concept-enhanced features (ablation) | EC-SL with MIL and ISAB · ActivityNet Captions validation | CIDEr: 20.98 -> 21.44 | Improved |
| add localizer-derived word-importance weights to the caption loss (ablation) | EC-SL with MIL, ISAB, and gating · ActivityNet Captions validation | CIDEr: 21.44 -> 21.21 | Hurt |
| vary induced set size from 8 to 16, 32, and 64 (ablation) | EC-SL · ActivityNet Captions validation | CIDEr: 21.05; 21.21; 21.44; 21.45 | Mixed |
| select inducing points from maximum-contribution concept features instead of randomly (ablation) | EC-SL · ActivityNet Captions validation | CIDEr: 20.67 -> 21.21 | Improved |
| use word-prediction logits rather than LSTM hidden states as sentence features (ablation) | EC-SL sentence localizer · ActivityNet Captions validation | CIDEr: 21.21 -> 18.41 | Hurt |
| use word embeddings rather than LSTM hidden states as sentence features (ablation) | EC-SL sentence localizer · ActivityNet Captions validation | CIDEr: 21.21 -> 21.36 | No clear effect |
Closest papers by meaning
- Weakly Supervised Dense Event Captioning in Videos (NeurIPS 2018 · similarity 0.95)
- Jointly Localizing and Describing Events for Dense Video Captioning (CVPR 2018 · similarity 0.92)
- Implicit Location-Caption Alignment via Complementary Masking for Weakly-Supervised Dense Video Captioning (AAAI 2025 · similarity 0.91)
- Event-Equalized Dense Video Captioning (CVPR 2025 · similarity 0.89)
- Exploring Temporal Event Cues for Dense Video Captioning in Cyclic Co-Learning (AAAI 2025 · similarity 0.89)
- Do You Remember? Dense Video Captioning with Cross-Modal Memory Retrieval (CVPR 2024 · 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.
