Research map / Video / Action recognition
Actor-centric Relation Network
Chen Sun, Abhinav Shrivastava, Carl Vondrick, Kevin Murphy, Rahul Sukthankar, Cordelia Schmid
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 |
|---|---|---|---|
| add ACRN relation module (ablation) | Base-Model · JHMDB (3 splits) | frame-AP: 75.2 -> 77.6 | Improved |
| use Resize+Concat of global features instead of Base-Model only (ablation) | Base-Model · JHMDB (3 splits) | frame-AP: 75.2 -> 74.8 | No clear effect |
| add Santoro et al. average-pooled relation network (ablation) | Base-Model · JHMDB (3 splits) | frame-AP: 75.2 -> 75.1 | No clear effect |
| add ACRN on JHMDB at 20 frames | Base-Model · JHMDB (3 splits) | video-AP: 78.8 -> 80.1 | Improved |
| add ACRN on AVA v2.1 | Base-Model · AVA v2.1 | frame-AP: 15.5 -> 17.4 | Improved |
| add ACRN with 1 input frame (ablation) | Base-Model · JHMDB (3 splits) | frame-AP: 52.6 -> 54.0 | Improved |
| add ACRN with 5 input frames (ablation) | Base-Model · JHMDB (3 splits) | frame-AP: 66.1 -> 69.8 | Improved |
| add ACRN with 10 input frames (ablation) | Base-Model · JHMDB (3 splits) | frame-AP: 70.6 -> 74.9 | Improved |
| increase input frames from 1 to 20 (ablation) | ACRN · JHMDB (3 splits) | frame-AP: 54.0 -> 77.6 | Improved |
| increase input frames from 1 to 20 (ablation) | Base-Model · JHMDB (3 splits) | frame-AP: 52.6 -> 75.2 | Improved |
| use Conv2c instead of Mixed_4f as global feature (ablation) | ACRN · JHMDB (3 splits), scale 1.0 | frame-AP: 77.6 -> 77.9 | Improved |
| combine Conv2c, Mixed_3c, Mixed_4f instead of Conv2c alone (ablation) | ACRN · JHMDB (3 splits), scale 1.0 | frame-AP: 77.9 -> 77.5 | Hurt |
Closest papers by meaning
- Actor-Context-Actor Relation Network for Spatio-Temporal Action Localization (CVPR 2021 · similarity 0.92)
- Relational Action Forecasting (CVPR 2019 · similarity 0.87)
- Action Genome: Actions As Compositions of Spatio-Temporal Scene Graphs (CVPR 2020 · similarity 0.87)
- Relational Prototypical Network for Weakly Supervised Temporal Action Localization (AAAI 2020 · similarity 0.87)
- CycleACR: Cycle Modeling of Actor-Context Relations for Video Action Detection (TPAMI 2025 · similarity 0.87)
- Unified Graph Structured Models for Video Understanding (ICCV 2021 · similarity 0.87)
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.
