Research map / Video / Action recognition
Learning to Recognize Actions on Objects in Egocentric Video With Attention Dictionaries
Swathikiran Sudhakaran, Sérgio Escalera, Oswald Lanz
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 recurrent attention (ablation) | ConvLSTM · EPIC-KITCHENS val (R2D-34) | action accuracy (top-1): 9.87 -> 12.25 | Improved |
| add fine-grained output gating (ablation) | ConvLSTM · EPIC-KITCHENS val (R2D-34) | action accuracy (top-1): 9.87 -> 13.75 | Improved |
| add recurrent attention and fine-grained output gating together (ablation) | ConvLSTM · EPIC-KITCHENS val (R2D-34) | action accuracy (top-1): 9.87 -> 14.92 | Improved |
| add attention on memory (full LSTA) (ablation) | ConvLSTM + rca-attn + gating · EPIC-KITCHENS val (R2D-34) | action accuracy (top-1): 14.92 -> 15.09 | Improved |
| increase memory attention dictionary size from 100 to 300 (ablation) | LSTA · EPIC-KITCHENS val (R2D-34) | action accuracy (top-1): 13.84 -> 15.09 | Improved |
| increase memory attention dictionary size from 300 to 400 (ablation) | LSTA · EPIC-KITCHENS val (R2D-34) | action accuracy (top-1): 15.09 -> 14.07 | Hurt |
| add object-centric stream (ablation) | LSTA · EPIC-KITCHENS val (R2D-34) | action accuracy (top-1): 15.09 -> 15.55 | Improved |
| add scene context stream (ablation) | LSTA · EPIC-KITCHENS val (R2D-34) | action accuracy (top-1): 15.09 -> 15.3 | Improved |
| add both object and context streams (ablation) | LSTA · EPIC-KITCHENS val (R2D-34) | action accuracy (top-1): 15.09 -> 16.63 | Improved |
| add bias control (action logits to verb/noun classifiers) (ablation) | EgoACO · EPIC-KITCHENS val (R2D-34) | noun accuracy (top-1): 10.22 vs 22.78 | Improved |
| use class activation pooling instead of attentional pooling (ablation) | EgoACO · EPIC-KITCHENS val (R2D-34) | action accuracy (top-1): 13.86 -> 16.63 | Improved |
| use class activation pooling instead of CAM (ablation) | EgoACO · EPIC-KITCHENS val (R2D-34) | action accuracy (top-1): 13.07 -> 16.63 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- LSTA — LSTA: Long Short-Term Attention for Egocentric Action Recognition (CVPR 2019) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Interactive Prototype Learning for Egocentric Action Recognition (ICCV 2021 · similarity 0.90)
- LSTA: Long Short-Term Attention for Egocentric Action Recognition (CVPR 2019 · similarity 0.89)
- Video Representation Learning Using Discriminative Pooling (CVPR 2018 · similarity 0.89)
- Action Scene Graphs for Long-Form Understanding of Egocentric Videos (CVPR 2024 · similarity 0.88)
- Symbiotic Attention with Privileged Information for Egocentric Action Recognition (AAAI 2020 · similarity 0.88)
- Egocentric Action Recognition by Automatic Relation Modeling (TPAMI 2022 · 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.
