Research map / 3D vision / Face and shape
Sparse to Dense Dynamic 3D Facial Expression Generation
Naima Otberdout, Claudio Ferrari, Mohamed Daoudi, Stefano Berretti, Alberto Del Bimbo
Check 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 S2D-Dec instead of PCA-220 | 3D facial expression reconstruction · CoMA expression split | mean per-vertex error (mm): 0.76 \pm 0.73 -> 0.52 \pm 0.59 | Improved |
| Use S2D-Dec instead of PCA-220 | 3D facial expression reconstruction · D3DFACS expression split | mean per-vertex error (mm): 0.42 \pm 0.44 -> 0.28\pm 0.31 | Improved |
| Use S2D-Dec instead of PCA-220 | 3D facial expression reconstruction · CoMA identity split | mean per-vertex error (mm): 0.80 \pm 0.73 -> 0.55 \pm 0.62 | Improved |
| Use S2D-Dec instead of PCA-220 | 3D facial expression reconstruction · D3DFACS identity split | mean per-vertex error (mm): 0.56 \pm 0.56 -> 0.27 \pm 0.30 | Improved |
| Generate landmark motion with Motion3DGAN instead of LSTM | CoMA dynamic expression generation · CoMA expression classification, Gen-LM | classification accuracy (%): 46.53 -> 65.28 | Improved |
| Generate landmark motion with Motion3DGAN instead of LSTM | CoMA dynamic expression generation · CoMA expression classification, Det-LM | classification accuracy (%): 52.08 -> 69.44 | Improved |
| Generate landmark motion with Motion3DGAN instead of LSTM | CoMA dynamic expression generation · CoMA expression generation, Gen-LM | FID: 21.76 -> 20.45 | Improved |
| Generate landmark motion with Motion3DGAN instead of LSTM | CoMA dynamic expression generation · CoMA expression generation, Det-LM | FID: 27.96 -> 19.01 | Improved |
| Add standard L1 reconstruction loss without landmark-distance weights (ablation) | S2D-Dec · CoMA reconstruction ablation | mean per-vertex error (mm): 1.27 \pm 1.88 -> 0.92 \pm 1.33 | Improved |
| Add landmark-distance weights to the L1 reconstruction loss (ablation) | S2D-Dec · CoMA reconstruction ablation | mean per-vertex error (mm): 0.92 \pm 1.33 -> 0.50\pm 0.56 | Improved |
Closest papers by meaning
- Learning to Generate Customized Dynamic 3D Facial Expressions (ECCV 2020 · similarity 0.92)
- Dynamic Facial Expression Generation on Hilbert Hypersphere With Conditional Wasserstein Generative Adversarial Nets (TPAMI 2020 · similarity 0.90)
- DEEPTalk: Dynamic Emotion Embedding for Probabilistic Speech-Driven 3D Face Animation (AAAI 2025 · similarity 0.90)
- 3D Facial Expressions through Analysis-by-Neural-Synthesis (CVPR 2024 · similarity 0.89)
- TEASER: Token Enhanced Spatial Modeling for Expressions Reconstruction (ICLR 2025 · similarity 0.88)
- OFER: Occluded Face Expression Reconstruction (CVPR 2025 · 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.
