Research map / Vision-language and multimodal / Human and embodied modeling
AURA: Visually Interpretable Affective Understanding via Robust Archetypes
Guanyu Hu, Dimitrios Kollias, Xinyu Yang
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 |
|---|---|---|---|
| replace fine-tuned CLIP with frozen CLIP plus AURA | CLIP · RAF-DB | accuracy: 89.1 -> 94.0 (+4.9) | Improved |
| replace fine-tuned CLIP with frozen CLIP plus AURA | CLIP · AffectNet-VA | CCC: 66.4 -> 74.1 (+7.7) | Improved |
| replace fine-tuned CLIP with frozen CLIP plus AURA | CLIP · EmotioNet | AU F1: 62.3 -> 67.3 (+5.0) | Improved |
| replace fine-tuned DINO with frozen DINO plus AURA | DINO · RAF-DB | accuracy: 88.3 -> 92.6 (+4.3) | Improved |
| replace fine-tuned DINO with frozen DINO plus AURA | DINO · AffectNet-VA | CCC: 65.4 -> 72.0 (+6.6) | Improved |
| replace fine-tuned DINO with frozen DINO plus AURA | DINO · EmotioNet | AU F1: 60.7 -> 65.4 (+4.7) | Improved |
| add visual archetypes instead of learned text prompts (CoOp) | CLIP · AffectNet-VA | CCC-VA: 44.5 -> 74.1 | Improved |
| vary Kmax over 150, 200, 300, 400 (ablation) | AURA · RAF-DB | accuracy: accuracy varies by at most 0.2% | No clear effect |
| add the VAS Projector (ablation) | AURA · all tasks | performance | Improved |
| add the Archetype Contextualization module (ablation) | AURA · all tasks | performance | Improved |
| add the Topology Refinement objective (ablation) | AURA · all tasks | performance | Improved |
| change loss weights lambda from uniform setting 1 (ablation) | AURA · all tasks | accuracy | No clear effect |
Closest papers by meaning
- Customizing Visual Emotion Evaluation for MLLMs: An Open-vocabulary, Multifaceted, and Scalable Approach (ICLR 2026 · similarity 0.88)
- A Unified and Interpretable Emotion Representation and Expression Generation (CVPR 2024 · similarity 0.86)
- Interpreting and Enhancing Emotional Circuits in Large Vision-Language Models via Cross-Modal Information Flow (ICML 2026 · similarity 0.85)
- Bridge Then Begin Anew: Generating Target-Relevant Intermediate Model for Source-Free Visual Emotion Adaptation (AAAI 2025 · similarity 0.85)
- EmoEdit: Evoking Emotions through Image Manipulation (CVPR 2025 · similarity 0.85)
- EmoVIT: Revolutionizing Emotion Insights with Visual Instruction Tuning (CVPR 2024 · similarity 0.85)
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.
