Research map / Label-efficient and robust learning / Recognition and long-tail
Φeat: Physically-Grounded Material Feature Representation
Giuseppe Vecchio, Adrien Kaiser, Claudia Cuttano, ROUFFET Romain, Rosalie MARTIN, Elena Garces, Tamy Boubekeur
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 |
|---|---|---|---|
| add single-render material supervision to DINOv3 (ablation) | DINOv3 · material k-NN classification | accuracy: 66.9 -> 34.5 | Hurt |
| add single-render material supervision to DINOv3 (ablation) | DINOv3 · DuMaS material selection | mIoU: 59.7 -> 69.4 | Improved |
| add multi-render scheme (ablation) | DINOv3 + single render · material k-NN classification | accuracy: 34.5 -> 51.3 | Improved |
| add multi-render scheme (ablation) | DINOv3 + single render · DuMaS material selection | F1: 78.1 -> 79.2 | Improved |
| add contrastive objective (ablation) | DINOv3 + multi render · material k-NN classification | accuracy: 51.3 -> 71.8 | Improved |
| add contrastive objective (ablation) | DINOv3 + multi render · DuMaS material selection | mIoU: 69.9 -> 72.0 | Improved |
| material pretraining Φeat vs DINOv3 | DINOv3 · ImageNet linear probe | accuracy: 83.5 -> 80.1 | Hurt |
| material pretraining Φeat vs DINOv3 | DINOv3 · ADE20K linear probe | mIoU: 51.5 -> 46.2 | Hurt |
| material pretraining Φeat vs DINOv3 | DINOv3 · Illumination robustness | pairwise Hamming distance: 0.240 -> 0.221 | Improved |
| material pretraining Φeat vs DINOv3 | DINOv3 · Geometry robustness | pairwise Hamming distance: 0.365 -> 0.305 | Improved |
Closest papers by meaning
- Alchemist: Parametric Control of Material Properties with Diffusion Models (CVPR 2024 · similarity 0.86)
- MaterialRefGS: Reflective Gaussian Splatting with Multi-view Consistent Material Inference (NeurIPS 2025 · similarity 0.86)
- MaRI: Material Retrieval Integration across Domains (CVPR 2025 · similarity 0.85)
- Intrinsic Image Diffusion for Indoor Single-view Material Estimation (CVPR 2024 · similarity 0.85)
- Material Palette: Extraction of Materials from a Single Image (CVPR 2024 · similarity 0.85)
- Material Anything: Generating Materials for Any 3D Object via Diffusion (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.
