Research map / Video / Action recognition
Cross-Task Weakly Supervised Learning From Instructional Videos
Check 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 |
|---|---|---|---|
| share component classifiers across primary and related tasks instead of training independent step models | CrossTask step recognition · CrossTask primary tasks test set | recall (%): 18.6 -> 22.4 | Helped |
| use component-level rather than step-level sharing, with primary and related tasks (ablation) | CrossTask step recognition · CrossTask primary tasks test set | recall (%): 19.8 -> 22.4 | Helped |
| add related tasks to component-based training (ablation) | CrossTask step recognition · CrossTask primary tasks test set | recall (%): 20.2 -> 22.4 | Helped |
| add component sharing across primary tasks (ablation) | CrossTask step recognition · CrossTask primary tasks test set | recall (%): 18.7 -> 20.2 | Helped |
| replace component-level sharing with step-level sharing (ablation) | CrossTask step recognition · CrossTask primary tasks test set | recall (%): 22.4 -> 19.8 | Hurt |
| remove narration-based temporal text constraints during training (ablation) | Proposed cross-task model · CrossTask primary tasks test set | recall (%): 22.4% -> 17% | Hurt |
| use learned wikiHow text aggregation instead of TF-IDF for narration localization (ablation) | Narration constraint localization · 18 primary tasks | precision (%): 13.3 -> 15.9 | Helped |
| use learned wikiHow text aggregation instead of max-pooled word vectors (ablation) | Narration constraint localization · 18 primary tasks | recall (%): 10.4 -> 13.9 | Helped |
| apply component sharing to cut shelf across dissimilar cooking and home-improvement contexts | CrossTask component model · cut shelf | recall (%): 47% -> 28% | Hurt |
| increase the number of related training tasks | CrossTask component model · Unseen primary tasks | recall | Helped |
Closest papers by meaning
- Unsupervised Learning From Narrated Instruction Videos (CVPR 2016 · similarity 0.88)
- Collaborative Weakly Supervised Video Correlation Learning for Procedure-Aware Instructional Video Analysis (AAAI 2024 · similarity 0.86)
- Instruction Embedding: Latent Representations of Instructions Towards Task Identification (NeurIPS 2024 · similarity 0.85)
- Masked Temporal Interpolation Diffusion for Procedure Planning in Instructional Videos (ICLR 2025 · similarity 0.85)
- Detours for Navigating Instructional Videos (CVPR 2024 · similarity 0.85)
- SLVMBench: Skill Learning from Video Memory (NeurIPS 2026 · 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.