Research map / Video / Video-language understanding
Aligning Step-by-Step Instructional Diagrams to Video Demonstrations
Jiahao Zhang, Anoop Cherian, Yanbin Liu, Yizhak Ben-Shabat, Cristian Rodriguez, Stephen Gould
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 |
|---|---|---|---|
| use the proposed three-loss contrastive framework instead of CLIP | video-diagram alignment · IAW step diagrams | Top1 Acc.%: 19.61 -> 28.62 | Improved |
| use the proposed framework instead of CLIP | video-diagram alignment · IAW step diagrams | R@1: 16.94 -> 22.30 | Improved |
| add sinusoidal progress rate features (SPRF) (ablation) | Ours · IAW step diagrams | Top1 Acc.%: 21.73 -> 28.62 | Improved |
| add sinusoidal progress rate features (SPRF) (ablation) | Ours · IAW step diagrams | AIE: 6.018 -> 3.734 | Improved |
| apply DTW instead of no post-processing (ablation) | Ours · IAW step diagrams | R@3: 45.00 -> 32.45 | Hurt |
| apply OT set matching (ablation) | Ours · IAW step diagrams | R@1: 22.30 -> 26.62 | Improved |
| use Video-Diagram loss A instead of standard CLIP infoNCE (ablation) | contrastive alignment model · IAW step diagrams | Top1 Acc.%: 19.61 -> 20.87 | Improved |
| use Video-Manual loss B (ablation) | contrastive alignment model · IAW step diagrams | Top1 Acc.%: 19.61 -> 27.20 | Improved |
| combine Video-Manual loss B with Video-Diagram loss A (ablation) | contrastive alignment model · IAW step diagrams | Top1 Acc.%: 27.20 -> 28.20 | Improved |
| add Intra-Manual loss C to losses A and B (ablation) | contrastive alignment model · IAW step diagrams | Top1 Acc.%: 28.20 -> 28.43 | Mixed |
| include page-diagram losses in addition to step-diagram losses (ablation) | loss configuration D1 · IAW step diagrams | Top1 Acc.%: 28.62 -> 28.26 | No clear effect |
| use page diagrams as training supervision in addition to step diagrams (ablation) | loss configuration D1 · IAW page diagrams | Top1 Acc.%: 33.72 -> 34.94 | Improved |
Closest papers by meaning
- IKEA Manuals at Work: 4D Grounding of Assembly Instructions on Internet Videos (NeurIPS 2024 · similarity 0.88)
- Aligning Videos in Space and Time (ECCV 2020 · similarity 0.87)
- Collaborative Weakly Supervised Video Correlation Learning for Procedure-Aware Instructional Video Analysis (AAAI 2024 · similarity 0.85)
- ShowHowTo: Generating Scene-Conditioned Step-by-Step Visual Instructions (CVPR 2025 · similarity 0.85)
- Step Differences in Instructional Video (CVPR 2024 · similarity 0.85)
- Instruction-based Image Manipulation by Watching How Things Move (CVPR 2025 · similarity 0.85)
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.
