Research map / Vision-language and multimodal / Human and embodied modeling
AutoAD: Movie Description in Context
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 multiple frames rather than one frame (ablation) | AutoAD · MAD-eval-Named | CIDEr: 4.0 -> 6.7 | Helped |
| Add previous AD context (ablation) | AutoAD · MAD-eval-Named | CIDEr: 6.7 -> 12.6 | Helped |
| Add AudioVault-AD text-only pretraining (ablation) | AutoAD with recurrent AD context · MAD-eval-Named | CIDEr: 12.6 -> 14.1 | Helped |
| Add WebVid visual pretraining after AudioVault-AD pretraining (ablation) | AutoAD · MAD-eval-Named, oracle context | CIDEr: 19.0 -> 21.9 | Helped |
| Add full subtitle context (ablation) | AutoAD with visual and previous-AD context · MAD-eval-Named | CIDEr: 14.3 -> 13.3 | Hurt |
| Filter subtitle context to character names (ablation) | AutoAD with visual and previous-AD context · MAD-eval-Named | CIDEr: 13.3 -> 14.2 | Helped |
| Train on denoised MAD-v2 instead of MAD-v1 (ablation) | AutoAD · MAD-eval-Named | CIDEr: 17.0 -> 17.8 | Helped |
| Use raw AD text tokens rather than mapped language features for context (ablation) | AutoAD · MAD-eval-Named | CIDEr: about 8.0 -> 12.6 | Helped |
| Add LSMDC training data to AutoAD trained on MAD-v2 Unnamed (ablation) | AutoAD · LSMDC 2019 Multi-Sentence Description public test | CIDEr: 16.7 -> 17.5 | Helped |
| Use AutoAD rather than ClipCap for movie AD generation | Movie AD generation · MAD-eval-Named | CIDEr: 4.4 -> 14.1 | Helped |
| Use AutoAD rather than CapDec with AudioVault-AD pretraining | Movie AD generation · MAD-eval-Named | CIDEr: 6.7 -> 14.1 | Helped |
Closest papers by meaning
- DistinctAD: Distinctive Audio Description Generation in Contexts (CVPR 2025 · similarity 0.92)
- AutoAD III: The Prequel - Back to the Pixels (CVPR 2024 · similarity 0.92)
- Contextual AD Narration with Interleaved Multimodal Sequence (CVPR 2025 · similarity 0.90)
- MM-Narrator: Narrating Long-form Videos with Multimodal In-Context Learning (CVPR 2024 · similarity 0.86)
- CALVIN: Improved Contextual Video Captioning via Instruction Tuning (NeurIPS 2024 · similarity 0.85)
- MICap: A Unified Model for Identity-Aware Movie Descriptions (CVPR 2024 · similarity 0.84)
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.