Research map / Architectures and efficiency / CNN design
Deep Affect Prediction in-the-Wild: Aff-Wild Database and Challenge, Deep Architectures, and Beyond
Dimitrios Kollias, Panagiotis Tzirakis, Mihalis A. Nicolaou, Athanasios Papaioannou, Guoying Zhao, Björn Wolfgang Schuller, Irene Kotsia, Stefanos P. Zafeiriou
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 |
|---|---|---|---|
| use mean of annotators instead of a single annotator as label (ablation) | VGG-Face · Aff-Wild test | CCC valence: 0.39 -> 0.51 | Improved |
| add 68 facial landmark inputs (ablation) | VGG-Face · Aff-Wild test | CCC valence: 0.44 -> 0.51 | Improved |
| add 68 facial landmark inputs (ablation) | VGG-Face · Aff-Wild test | MSE valence: 0.12 -> 0.10 | Improved |
| use VGG-Face pretraining instead of VGG-16 (ablation) | CNN · Aff-Wild test | CCC mean: 0.35 -> 0.42 | Improved |
| use VGG-Face pretraining instead of ResNet-50 (ablation) | CNN · Aff-Wild test | CCC mean: 0.37 -> 0.42 | Improved |
| add GRU RNN on top of CNN (2 layers, 128 units) (ablation) | VGG-Face · Aff-Wild test | CCC mean: 0.42 -> 0.50 | Improved |
| use GRU instead of LSTM in the RNN part (ablation) | VGG-Face CNN-RNN · Aff-Wild test | CCC mean: 0.45 -> 0.50 | Improved |
| use 2 RNN layers instead of 1 (128 units) (ablation) | VGG-Face-GRU · Aff-Wild test | CCC valence: 0.53 -> 0.57 | Improved |
| use 128 hidden units instead of 100 (2 layers) (ablation) | VGG-Face-GRU · Aff-Wild test | CCC valence: 0.50 -> 0.57 | Improved |
| pretrain on Aff-Wild before fine-tuning on RECOLA (ablation) | AffWildNet · RECOLA test | CCC valence: 0.462 -> 0.526 | Improved |
| fine-tune AffWildNet instead of training CNN only on AFEW-VA (ablation) | AffWildNet · AFEW-VA | CCC arousal: 0.474 -> 0.556 | Improved |
| fine-tune AffWildNet for categorical emotion instead of VGG-Face + RNN (ablation) | AffWildNet · EmotiW 2017 val | total accuracy: 0.384 -> 0.454 | Improved |
Closest papers by meaning
- FEEL: Quantifying Heterogeneity in Physiological Signals for Generalizable Emotion Recognition (NeurIPS 2025 · similarity 0.90)
- AffectGPT: A New Dataset, Model, and Benchmark for Emotion Understanding with Multimodal Large Language Models (ICML 2025 · similarity 0.88)
- OV-MER: Towards Open-Vocabulary Multimodal Emotion Recognition (ICML 2025 · similarity 0.87)
- AffectGPT-RL: Revealing Roles of Reinforcement Learning in Open-Vocabulary Emotion Recognition (NeurIPS 2026 · similarity 0.86)
- EmoNet-Face: An Expert-Annotated Benchmark for Synthetic Emotion Recognition (NeurIPS 2025 · similarity 0.86)
- FindingEmo: An Image Dataset for Emotion Recognition in the Wild (NeurIPS 2024 · similarity 0.86)
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.
