Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Hierarchical Material Recognition from Local Appearance
Matthew Beveridge, Shree K. Nayar
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 |
|---|---|---|---|
| hierarchical graph-attention classifier instead of FRP | ResNet50 · FMD | top-1 accuracy: 88.8 -> 96.1 | Improved |
| hierarchical graph-attention classifier instead of FRP | ResNet50 · KTH-TIPS-2b | top-1 accuracy: 90.7 -> 93.5 | Improved |
| hierarchical graph classifier instead of CLASSNet | ResNet50 · GTOS | top-1 accuracy: 85.6 -> 87.9 | Improved |
| hierarchical graph classifier instead of DSRNet | ResNet50 · GTOS-Mobile | top-1 accuracy: 87.0 -> 92.2 | Improved |
| hierarchical model instead of best end-to-end baseline (DEPNet) | ResNet50 · Matador-C1 OOD | top-1 accuracy: 76.1 -> 82.9 | Improved |
| hierarchical model instead of best end-to-end baseline (MSLac) | ResNet50 · Matador-C1 grayscale | top-1 accuracy: 84.7 -> 87.5 | Improved |
| hierarchical model instead of FRP | ResNet50 · Matador-C1 color | top-1 accuracy: 89.4 -> 94.1 | Improved |
| finetune end-to-end instead of classifier head only | MSLac · Matador-C1 color | top-1 accuracy: 79.1 -> 88.5 | Improved |
| finetune end-to-end instead of classifier head only | DeepTEN · Matador-C1 color | top-1 accuracy: 54.0 -> 88.8 | Improved |
| add rendered novel views to training data (ablation) | hierarchical model · Matador OOD test set | top-1 accuracy: up to 4.9% | Improved |
| reintroduce 16 samples of a held-out class (few-shot) | hierarchical model · Matador-C1 held-out class | accuracy: ~90% accuracy with 16 samples | Improved |
Closest papers by meaning
- Harnessing the Power of Foundation Models for Accurate Material Classification (CVPR 2026 · similarity 0.89)
- One-Shot Recognition of Any Material Anywhere Using Contrastive Learning with Physics-Based Rendering (ICCV 2023 · similarity 0.89)
- Recognizing Material Properties from Images (TPAMI 2019 · similarity 0.88)
- Infusing Synthetic Data with Real-World Patterns for Zero-Shot Material State Segmentation (NeurIPS 2024 · similarity 0.87)
- MaRI: Material Retrieval Integration across Domains (CVPR 2025 · similarity 0.85)
- Intrinsic Image Diffusion for Indoor Single-view Material Estimation (CVPR 2024 · similarity 0.84)
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.
