Research map / Video / Action recognition
Spatial-Aware Object Embeddings for Zero-Shot Localization and Classification of Actions
Pascal Mettes, Cees G. M. Snoek
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 relevant local objects to actor-only scoring (ablation) | actor-only embedding · UCF Sports | localization mAP @0.5: 0.083 -> 0.175 | Improved |
| add spatial relations between actors and objects (ablation) | actors and objects embedding (1 object) · UCF Sports | localization mAP @0.5: 0.175 -> 0.221 | Improved |
| add spatial relations (5 objects) (ablation) | actors and objects embedding · UCF Sports | classification accuracy: 0.139 -> 0.255 | Improved |
| add spatial relations (1 object) (ablation) | actors and objects embedding · UCF Sports | classification accuracy: 0.205 -> 0.180 | Hurt |
| use more local objects per action (ablation) | spatial-aware embedding · UCF Sports | localization mAP @0.5: 0.221 -> 0.199 | Hurt |
| use more local objects per action (ablation) | spatial-aware embedding · UCF Sports | classification accuracy: 0.180 -> 0.255 | Improved |
| use finer spatial grid (4x4 or 5x5) instead of 3x3 (ablation) | spatial-aware embedding · UCF Sports | localization mAP @0.5: 0.199 -> 0.170 | Hurt |
| add global object classifier scores (ablation) | spatial-aware embedding · UCF Sports | classification accuracy: 0.255 -> 0.645 | Improved |
| add global object classifier scores (ablation) | spatial-aware embedding · UCF Sports | localization AUC | Improved |
| use smaller global network instead of 13k-category GoogLeNet (ablation) | spatial- and global-aware embedding · UCF Sports | classification accuracy: 0.645 -> 0.374 | Hurt |
Closest papers by meaning
- Object Priors for Classifying and Localizing Unseen Actions (IJCV 2021 · similarity 0.93)
- Universal Prototype Transport for Zero-Shot Action Recognition and Localization (IJCV 2023 · similarity 0.90)
- Graph Convolutional Module for Temporal Action Localization in Videos (TPAMI 2021 · similarity 0.89)
- Pointly-Supervised Action Localization (IJCV 2018 · similarity 0.89)
- Watch Only Once: An End-to-End Video Action Detection Framework (ICCV 2021 · similarity 0.88)
- VideoGEM: Training-free Action Grounding in Videos (CVPR 2025 · 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.
