Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
RAID-G: Robust Estimation of Approximate Infinite Dimensional Gaussian With Application to Material Recognition
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 RAID-G with the χ² kernel mapping instead of FV-CNN | Material recognition with VGG-VD19 features · FMD | accuracy: 79.8 ± 1.8 -> 84.9 ± 1.4 | Helped |
| Use RAID-G with the χ² kernel mapping instead of FV-CNN | Material recognition with VGG-VD19 features · UIUC material | accuracy: 80.5 ± 2.7 -> 86.3 ± 2.9 | Helped |
| Use RAID-G with the χ² kernel mapping instead of FV-CNN | Material recognition with VGG-VD19 features · KTH-TIPS 2b | accuracy: 81.8 ± 2.5 -> 81.3 ± 1.6 | Hurt |
| Use RAID-G with the χ² kernel mapping instead of FC + FV-CNN | Material recognition with VGG-VD19 features · DTD | accuracy: 74.7 ± 1.0 -> 76.4 ± 1.1 | Helped |
| Use RAID-G with the χ² kernel mapping instead of FC + FV-CNN | Material recognition with VGG-VD19 features · Open Surfaces | accuracy: 60.9 -> 61.1 | No clear effect |
| Add the von Neumann regularized MLE estimator instead of the Ledoit-Wolf estimator (ablation) | Gaussian descriptors with CNN features in the original space · FMD | accuracy: 81.3 ± 1.4 -> 83.6 ± 1.6 | Helped |
| Add the von Neumann regularized MLE estimator instead of the Ledoit-Wolf estimator (ablation) | Gaussian descriptors with CNN features in the original space · UIUC material | accuracy: 81.7 ± 2.9 -> 84.5 ± 1.8 | Helped |
| Use the χ² kernel mapping rather than the original feature space (ablation) | Gaussian descriptors with vN-MLE · FMD | accuracy: 83.6 ± 1.6 -> 84.9 ± 1.4 | Helped |
| Use the χ² kernel mapping rather than Hellinger's kernel mapping (ablation) | RAID-G-CNN · FMD | accuracy: 84.4 ± 1.3 -> 84.9 ± 1.4 | Helped |
| Use F-norm instead of Log-Euclidean metric (ablation) | RAID-G-CNN-Hel · FMD | accuracy: 82.3 ± 1.3 -> 84.4 ± 1.3 | Helped |
| Use F-norm instead of Log-Euclidean metric (ablation) | RAID-G-CNN-Hel · UIUC material | accuracy: 80.6 ± 4.3 -> 85.7 ± 2.1 | Helped |
| Replace the proposed mapping with random Fourier mapping (ablation) | RAID-G-CNN · FMD | accuracy: 84.9 ± 1.4 -> 79.7 ± 1.6 | Hurt |
Closest papers by meaning
- Approximate Log-Hilbert-Schmidt Distances Between Covariance Operators for Image Classification (CVPR 2016 · similarity 0.84)
- EigenGS Representation: From Eigenspace to Gaussian Image Space (CVPR 2025 · similarity 0.81)
- Robust Visual Recognition with Class-Imbalanced Open-World Noisy Data (AAAI 2024 · similarity 0.81)
- Selective Disk Bispectrum: A Complete and Rotation Invariant Image Descriptor (NeurIPS 2026 · similarity 0.80)
- Large Images Are Gaussians: High-Quality Large Image Representation with Levels of 2D Gaussian Splatting (AAAI 2025 · similarity 0.80)
- Mitigating the Curse of Dimensionality for Certified Robustness via Dual Randomized Smoothing (ICLR 2024 · similarity 0.80)
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.