Research map / Video / Action recognition
Actor-Centered Representations for Action Localization in Streaming Videos
Sathyanarayanan Aakur, Sudeep Sarkar
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 actor-centered hierarchical prediction instead of flat PredLearn | PredLearn · UCF Sports | video mAP @0.5: 0.32 -> 0.59 | Improved |
| add actor-centered hierarchical prediction instead of flat PredLearn | PredLearn · JHMDB | video mAP @0.2: 0.30 -> 0.43 | Improved |
| add actor-centered hierarchical prediction instead of flat PredLearn | PredLearn · THUMOS'13 | video mAP @0.5: 0.10 -> 0.20 | Improved |
| add actor-centered hierarchical prediction instead of flat PredLearn | PredLearn · JHMDB | video mAP @0.5: 0.10 -> 0.15 | Improved |
| train on UCF Sports and test on JHMDB without finetuning, AC-HPL vs PredLearn | PredLearn · Cross-dataset generalization | mAP @0.2: 0.19 -> 0.39 | Improved |
| increase number of attention grids K from 1 to 25 (ablation) | AC-HPL · Collective Activities localization | mAP @0.5: 0.017 -> 0.449 | Improved |
| increase number of attention grids K from 1 to 25 (ablation) | AC-HPL · Collective Activities localization | mAP @0.2: 0.442 -> 0.723 | Improved |
| increase attention grids K from 1 to 25 (recall) (ablation) | AC-HPL · Collective Activities localization | recall: 0.181 -> 0.634 | Improved |
| fine-tune features with a 2-layer network instead of k-means labeling | AC-HPL · Collective Activities recognition | accuracy: 72.2 -> 80.2 | Improved |
| add both geometry prediction and contextualized feature prediction terms to the loss (ablation) | AC-HPL · UCF Sports | mAP | Improved |
| add hierarchical prediction with event-centric features (ablation) | AC-HPL · UCF Sports | mAP | Improved |
| use EdgeBox or Selective Search instead of SSD proposals (ablation) | AC-HPL · UCF Sports | mAP | No clear effect |
Closest papers by meaning
- Action Localization through Continual Predictive Learning (ECCV 2020 · similarity 0.89)
- Open-ended Hierarchical Streaming Video Understanding with Vision Language Models (ICCV 2025 · similarity 0.89)
- Temporal Action Localization in the Deep Learning Era: A Survey (TPAMI 2023 · similarity 0.88)
- Two-Stream Networks for Weakly-Supervised Temporal Action Localization With Semantic-Aware Mechanisms (CVPR 2023 · similarity 0.88)
- PivoTAL: Prior-Driven Supervision for Weakly-Supervised Temporal Action Localization (CVPR 2023 · similarity 0.88)
- Watch Only Once: An End-to-End Video Action Detection Framework (ICCV 2021 · similarity 0.88)
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.
