Research map / Label-efficient and robust learning / Recognition and long-tail
Confidence-Weighted Local Expression Predictions for Occlusion Handling in Expression Recognition and Action Unit Detection
Arnaud Dapogny, Kévin Bailly, Séverine Dubuisson
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 |
|---|---|---|---|
| make trees local (locality R=0.1) instead of global (ablation) | RS-RF · CK+ (8 expressions) | accuracy (%): 91.5 -> 93.4 | Improved |
| make trees local instead of global (ablation) | RS-RF · BU4D | accuracy (%): 73.1 -> 74.3 | Improved |
| add autoencoder confidence weighting to local predictions (ablation) | LS-RF · BU4D | accuracy (%): 74.3 -> 75.0 | Improved |
| add autoencoder confidence weighting to local predictions (ablation) | LS-RF · SFEW | accuracy (%): 35.6 -> 37.1 | Improved |
| add autoencoder confidence weighting to local predictions (ablation) | LS-RF · CK+ (8 expressions) | accuracy (%): 93.4 -> 93.4 | No clear effect |
| add local trees and confidence weighting under mouth occlusion (ablation) | global RS-RF · CK+ (8 expressions), mouth occluded | accuracy (%): 25.4 -> 67.1 | Improved |
| use WLS-RF instead of prior occlusion method under mouth occlusion | WLS-RF · Mouth occlusion | accuracy (%): 30.3 -> 67.1 | Improved |
| merge categorical datasets (M2) instead of CK+ only for LEP training (ablation) | AU detection · CK+ AUs | average AUC (%): 93.4 -> 93.7 | Improved |
| merge categorical datasets (M2) instead of CK+ only for LEP training (ablation) | AU detection · BP4D AUs | average AUC (%): 70.8 -> 70.6 | No clear effect |
| merge datasets (M1) instead of CK+ only (ablation) | AU detection · DISFA AUs | average AUC (%): 72.9 -> 75.9 | Improved |
| train LEPs on SFEW only instead of CK+ | AU detection · CK+ AUs | average AUC (%): 93.4 -> 88.9 | Hurt |
| use LEP features instead of SHTL | AU detection · CK+ AUs | average AUC (%): 91.7 -> 93.7 | Improved |
Closest papers by meaning
- Learning Deep Hierarchical Features with Spatial Regularization for One-Class Facial Expression Recognition (AAAI 2023 · similarity 0.88)
- Knowledge Augmented Deep Neural Networks for Joint Facial Expression and Action Unit Recognition (NeurIPS 2020 · similarity 0.88)
- Noise-Robust Facial Expression Recognition via Mamba-driven Neighbor Weight Refinement (ECCV 2026 · similarity 0.88)
- Open-Set Facial Expression Recognition (AAAI 2024 · similarity 0.88)
- Cross-Domain Facial Expression Recognition: A Unified Evaluation Benchmark and Adversarial Graph Learning (TPAMI 2021 · similarity 0.88)
- Navigating Label Ambiguity for Facial Expression Recognition in the Wild (AAAI 2025 · 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.
