Research map / Video / Video-language understanding
SODA: Story Oriented Dense Video Captioning Evaluation Framework
Soichiro Fujita, Tsutomu Hirao, Hidetaka Kamigaito, Manabu Okumura, Masaaki Nagata
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 |
|---|---|---|---|
| increase number of evaluated captions from m=1.0 to all | Current ActivityNet scorer (E2E Transformer) · ActivityNet Captions val | score (%): 4.10 -> 4.19 | No clear effect |
| increase number of evaluated captions from m=1.0 to all | Current ActivityNet scorer (LSTM) · ActivityNet Captions val | score (%): 4.92 -> 4.97 | No clear effect |
| increase number of evaluated captions from m=1.0 to all | SODA (c) (E2E Transformer) · ActivityNet Captions val | F1 (%): 4.02 -> 0.63 | Hurt |
| increase number of evaluated captions from m=1.0 to all | SODA (c) (LSTM) · ActivityNet Captions val | F1 (%): 3.15 -> 0.71 | Hurt |
| reduce number of captions from m=1.0 to m=0.1 | SODA (c) (E2E Transformer) · ActivityNet Captions val | F1 (%): 4.02 -> 1.47 | Hurt |
| use F-measure with tau 0 (b) instead of averaging over tau (a) (ablation) | SODA E2E Transformer · ActivityNet Captions val, m=1.0 | F1 (%): 3.86 -> 7.55 | Improved |
| add IoU-weighted METEOR cost (SODA c) instead of METEOR-only cost (SODA b) (ablation) | SODA LSTM · ActivityNet Captions val, m=1.0 | F1 (%): 6.07 -> 3.15 | No clear effect |
| shuffle caption order instead of correct order | SODA (c) (E2E Transformer) · oracle captions | score (%): 17.8 -> 7.66 | Hurt |
| shuffle caption order instead of correct order | Current ActivityNet scorer (E2E Transformer) · oracle captions | score (%): 16.1 -> 10.8 | Hurt |
| swap two adjacent captions instead of correct order | SODA (c) (LSTM) · oracle captions | score (%): 10.7 -> 8.89 | Hurt |
| swap two adjacent captions instead of correct order | Current ActivityNet scorer (LSTM) · oracle captions | score (%): 10.5 -> 9.92 | Hurt |
| use SODA (c) instead of current scorer for agreement with human ordering judgment | SODA (c) · manual evaluation, Swap vs Shuffle | accuracy: 0.72 -> 0.94 | Improved |
Closest papers by meaning
- HiCM²: Hierarchical Compact Memory Modeling for Dense Video Captioning (AAAI 2025 · similarity 0.87)
- VidChain: Chain-of-Tasks with Metric-based Direct Preference Optimization for Dense Video Captioning (AAAI 2025 · similarity 0.87)
- DIBS: Enhancing Dense Video Captioning with Unlabeled Videos via Pseudo Boundary Enrichment and Online Refinement (CVPR 2024 · similarity 0.87)
- Event-Equalized Dense Video Captioning (CVPR 2025 · similarity 0.86)
- G-VEval: A Versatile Metric for Evaluating Image and Video Captions Using GPT-4o (AAAI 2025 · similarity 0.86)
- A Picture is Worth More Than 77 Text Tokens: Evaluating CLIP-Style Models on Dense Captions (CVPR 2024 · 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.
