Research map / Video / Action recognition
Video-Mined Task Graphs for Keystep Recognition in Instructional Videos
Kumar Ashutosh, Santhosh Kumar Ramakrishnan, Triantafyllos Afouras, Kristen Grauman
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 probabilistic task graph instead of non-probabilistic (ablation) | PathSearch keystep assignment · COIN text-only | accuracy (%): 10.5 -> 16.3 | Improved |
| use fixed confidence threshold instead of adaptive per-video threshold (ablation) | PathSearch · COIN text-only | accuracy (%): 12.5 -> 16.3 | Improved |
| use fixed threshold instead of adaptive (ablation) | PathSearch · CrossTask text-only | accuracy (%): 17.1 -> 20.0 | Improved |
| use PathSearch (beam search) instead of Bayesian recursive filtering (ablation) | keystep assignment · COIN text-only | accuracy (%): 13.5 -> 16.3 | Improved |
| prioritize video keysteps over fusion for video-text (ablation) | keystep assignment · COIN video-text | accuracy (%): 14.0 -> 16.9 | Improved |
| lower text threshold gamma from 0.5 to 0.3 (ablation) | PathSearch · COIN text-only | accuracy (%): 16.3 -> 15.9 | Hurt |
| raise video threshold gamma from 0.3 to 0.5 (ablation) | PathSearch · COIN video-only | accuracy (%): 15.4 -> 14.9 | Hurt |
| add video-mined task graph to similarity baseline | Ours vs VideoCLIP · CrossTask video-only | accuracy (%): 28.5 -> 28.6 | No clear effect |
| add task graph under low keystep predictability (ablation) | Ours vs DistantSup. · COIN text-only | accuracy (%): 8.3 -> 14.1 | Improved |
| increase vocabulary with noisy keysteps to 10 x N | Ours vs DistantSup. · COIN text-only | accuracy (%): 4.3 -> 8.4 | Improved |
Closest papers by meaning
- Self-Supervised Multi-Task Procedure Learning from Instructional Videos (ECCV 2020 · similarity 0.87)
- Task Graph Maximum Likelihood Estimation for Procedural Activity Understanding in Egocentric Videos (TPAMI 2026 · similarity 0.87)
- Differentiable Task Graph Learning: Procedural Activity Representation and Online Mistake Detection from Egocentric Videos (NeurIPS 2024 · similarity 0.87)
- Learning Procedure-Aware Video Representation From Instructional Videos and Their Narrations (CVPR 2023 · similarity 0.87)
- StepFormer: Self-Supervised Step Discovery and Localization in Instructional Videos (CVPR 2023 · similarity 0.86)
- Learning from Narrated Instruction Videos (TPAMI 2017 · 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.
