Research map / 3D vision / Face and shape
SEREP: Semantic Facial Expression Representation for Robust In-the-Wild Capture and Retargeting
Arthur Josi, Luiz Gustavo Hafemann, Abdallah Dib, Emeline Got, Rafael M. O. Cruz, Marc-André Carbonneau
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 |
|---|---|---|---|
| remove the delta loss from the semantic expression model (ablation) | SEREP expression model · Expression reconstruction | mean reconstruction error (mm): 1.75 -> 1.15 | Improved |
| add the delta loss to the semantic expression model (ablation) | SEREP expression model · Identity preservation CSIM | CSIM: 0.736 -> 0.791 | Improved |
| condition the decoder on identity instead of adding expression displacements directly (ablation) | SEREP expression model · Identity preservation CSIM | CSIM: 0.773 -> 0.791 | Improved |
| condition the decoder on identity (ablation) | SEREP expression model · Expression reconstruction | mean reconstruction error (mm): 1.15 -> 1.15 | No clear effect |
| use SEREP semantic expression model instead of FLAME | expression model · Expression reconstruction | mean reconstruction error (mm): 1.36 -> 1.15 | Improved |
| use SEREP instead of FLAME for retargeting | expression model · Identity preservation CSIM | CSIM: 0.766 -> 0.791 | Improved |
| add real in-the-wild data to synthetic-only training (ablation) | SEREP capture model · MultiREX | average error (mm): 2.30 -> 2.19 | Improved |
| add the domain loss L_d (ablation) | SEREP capture model · MultiREX | average error (mm): 2.10 -> 2.19 | Hurt |
| add the domain loss L_d (ablation) | SEREP capture model · in-the-wild capture | expressiveness (qualitative) | Improved |
| use SEREP instead of EMICA across viewpoints | capture model · MultiREX Profile view | per-vertex error (mm): 2.49 -> 2.30 | Improved |
| use SEREP instead of SMIRK on profile views | capture model · MultiREX Profile view | per-vertex error (mm): 3.11 -> 2.30 | Improved |
| use SEREP instead of EMICA on forehead region | capture model · MultiREX Forehead | per-vertex error (mm): 1.65 -> 1.89 | Hurt |
Closest papers by meaning
- TEASER: Token Enhanced Spatial Modeling for Expressions Reconstruction (ICLR 2025 · similarity 0.87)
- EMOCA: Emotion Driven Monocular Face Capture and Animation (CVPR 2022 · similarity 0.87)
- 3D Facial Expressions through Analysis-by-Neural-Synthesis (CVPR 2024 · similarity 0.86)
- MimicME: A Large Scale Diverse 4D Database for Facial Expression Analysis (ECCV 2022 · similarity 0.86)
- SynFER: Towards Boosting Facial Expression Recognition with Synthetic Data (ICCV 2025 · similarity 0.86)
- Expressive Gaussian Human Avatars from Monocular RGB Video (NeurIPS 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.
