Research map / 3D vision / Face and shape
Cross-Modal Perceptionist: Can Face Geometry Be Gleaned From Voices?
Cho-Ying Wu, Chin-Cheng Hsu, Ulrich Neumann
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 |
|---|---|---|---|
| jointly train speech-to-image and image-to-3D prediction with voice information rather than cascade separately pretrained modules | CMP-supervised · Voxceleb-3D test | Mean ARE: 0.0302 -> 0.0241 | Improved |
| jointly train speech-to-image and image-to-3D prediction with voice information rather than cascade separately pretrained modules | CMP-unsupervised · Voxceleb-3D test | Mean ARE: 0.0302 -> 0.0251 | Improved |
| learn face geometry from voice using the supervised paired-voice/mesh framework | CMP-supervised · Voxceleb-3D test | ARE ER: 0.0311 -> 0.0152 | Improved |
| learn face geometry from voice using the supervised paired-voice/mesh framework | CMP-supervised · Voxceleb-3D test | ARE FR: 0.0173 -> 0.0186 | Hurt |
| learn face geometry from voice using the unsupervised joint framework | CMP-unsupervised · Voxceleb-3D test | ARE ER: 0.0311 -> 0.0181 | Improved |
| learn face geometry from voice using the unsupervised joint framework | CMP-unsupervised · Voxceleb-3D test | ARE MR: 0.0173 -> 0.0174 | Hurt |
| use RKD or PKT instead of vanilla KD (ablation) | CMP-unsupervised · Voxceleb-3D unsupervised | Mean ARE: 0.0298 -> 0.0254 | Improved |
| use attention-map similarity KD instead of vanilla KD (ablation) | CMP-unsupervised · Voxceleb-3D unsupervised | Mean ARE: 0.0298 -> 0.0304 | Hurt |
| use SP KD instead of vanilla KD (ablation) | CMP-unsupervised · Voxceleb-3D unsupervised | Mean ARE: 0.0298 -> 0.0262 | Improved |
| use Correlation KD instead of vanilla KD (ablation) | CMP-unsupervised · Voxceleb-3D unsupervised | Mean ARE: 0.0298 -> 0.0263 | Improved |
| use CRD instead of vanilla KD (ablation) | CMP-unsupervised · Voxceleb-3D unsupervised | Mean ARE: 0.0298 -> 0.0255 | Improved |
| use VID instead of vanilla KD (ablation) | CMP-unsupervised · Voxceleb-3D unsupervised | Mean ARE: 0.0298 -> 0.0259 | Improved |
Closest papers by meaning
- Seeing Voices and Hearing Faces: Cross-Modal Biometric Matching (CVPR 2018 · similarity 0.89)
- Cross-modal Associations in Vision and Language Models: Revisiting the Bouba-Kiki Effect (NeurIPS 2025 · similarity 0.85)
- FaceSpeak: Expressive and High-Quality Speech Synthesis from Human Portraits of Different Styles (AAAI 2025 · similarity 0.85)
- Perceptually Accurate 3D Talking Head Generation: New Definitions, Speech-Mesh Representation, and Evaluation Metrics (CVPR 2025 · similarity 0.85)
- FaceChain-ImagineID: Freely Crafting High-Fidelity Diverse Talking Faces from Disentangled Audio (CVPR 2024 · similarity 0.85)
- Let's Chorus: Partner-aware Hybrid Song-Driven 3D Head Animation (CVPR 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.
