Research map / Video / Action recognition
Semantically Guided Representation Learning For Action Anticipation
Anxhelo Diko, Danilo Avola, Bardh Prenkaj, Federico Fontana, Luigi Cinque
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 semantic prototype learning with cosine attention head (ablation) | baseline ViT-B anticipation model · EK100 val (RGB) | Action Top-5 recall: 15.2 -> 17.8 | Improved |
| add TCA block (ablation) | baseline ViT-B anticipation model · EK100 val (RGB) | Action Top-5 recall: 15.2 -> 16.7 | Improved |
| add TCA to PA + Sem to form full S-GEAR (ablation) | S-GEAR · EK100 val (RGB) | Action Top-5 recall: 18.0 -> 18.3 | Improved |
| learn visual prototypes instead of aligning to fixed language prototypes (ablation) | S-GEAR · EK100 val (RGB) | Action Top-5 recall: 17.4 -> 18.3 | Improved |
| use STSB sentence encoder instead of BERT for prototypes (ablation) | S-GEAR · EK100 | Top-5 recall | Improved |
| use only 10% of prototypes for relative positions (ablation) | S-GEAR · EK100 val (RGB) | Action Top-5 recall: 18.3 -> 17.8 | No clear effect |
| late-fuse two ViT-B models (224 and 384 input) | S-GEAR · EK100 val (RGB) | Action Top-5 recall: 18.3 -> 19.6 | Improved |
| late-fuse RGB and object modalities | S-GEAR · EK100 val | Action Top-5 recall: 18.9 -> 19.9 | Improved |
| increase anticipation time beyond training value | S-GEAR on EGTEA Gaze+ · EGTEA Gaze+ | Top-5 accuracy | Hurt |
| use ViT-B backbone vs TSN or irCSN features | S-GEAR · EK100 val (RGB) | Action Top-5 recall: 14.9 -> 18.3 | Improved |
Closest papers by meaning
- Prototypical Action Reasoning Facilitated by Vision-Language Alignment for Egocentric Action Anticipation (CVPR 2026 · similarity 0.91)
- Open-Vocabulary Long Term Action Anticipation (ECCV 2026 · similarity 0.90)
- Can't Make an Omelette Without Breaking Some Eggs: Plausible Action Anticipation Using Large Video-Language Models (CVPR 2024 · similarity 0.89)
- Towards Understanding Future: Consistency Guided Probabilistic Modeling for Action Anticipation (AAAI 2024 · similarity 0.88)
- AntGPT: Can Large Language Models Help Long-term Action Anticipation from Videos? (ICLR 2024 · similarity 0.88)
- Uncertainty-aware Action Decoupling Transformer for Action Anticipation (CVPR 2024 · 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.
