Research map / Architectures and efficiency / CNN design
Learning Partial Correlation Based Deep Visual Representation for Image Classification
Saimunur Rahman, Piotr Koniusz, Lei Wang, Luping Zhou, Peyman Moghadam, Changming Sun
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 |
|---|---|---|---|
| replace covariance pooling with sparse inverse-covariance pooling | ResNet-50 visual representation · Airplane | accuracy: 91.2 -> 92.7 | Improved |
| replace covariance pooling with sparse inverse-covariance pooling | ResNet-50 visual representation · Cars | accuracy: 92.1 -> 93.5 | Improved |
| replace covariance pooling with sparse inverse-covariance pooling | ResNet-50 visual representation · DTD | accuracy: 73.7 -> 60.7 | Hurt |
| replace covariance pooling with sparse inverse-covariance pooling | ResNet-50 visual representation · ImageNet100 | Top-1 accuracy: 71.5 -> 74.8 | Improved |
| increase matrix dimension from 256×256 to 512×512 (ablation) | VGG-16 iSICE · MIT Indoor | accuracy: 78.7 -> 81.1 | Improved |
| increase matrix dimension from 256×256 to 512×512 (ablation) | VGG-16 iSICE · Cars | accuracy: 94 -> 94.6 | Improved |
| add learning-rate and sparsity modulators (ablation) | iSICE with ResNet-50 · Airplane | accuracy: 92.7 -> 93.4 | Improved |
| add learning-rate and sparsity modulators (ablation) | iSICE with ResNet-50 · ImageNet100 | accuracy: 74.8 -> 76.3 | Improved |
| condition modulators on concatenated feature map and Δ1 rather than feature map alone (ablation) | iSICE+MLP with ResNet-50 · ImageNet100 | accuracy: 76.3 -> 77.1 | Improved |
| vary sparsity constant λ across the tested range (ablation) | iSICE · fine-grained datasets | accuracy: 92.56 ±0.25 | No clear effect |
| vary learning rate η across the tested range (ablation) | iSICE · Birds | accuracy: 86.51±0.14 | No clear effect |
| vary iteration count N (ablation) | iSICE · MIT Indoor | accuracy: 78.28; 78.66; 78.36 | No clear effect |
Closest papers by meaning
- Beyond Covariance: SICE and Kernel Based Visual Feature Representation (IJCV 2020 · similarity 0.88)
- On the Versatile Uses of Partial Distance Correlation in Deep Learning (ECCV 2022 · similarity 0.87)
- Is Second-Order Information Helpful for Large-Scale Visual Recognition? (ICCV 2017 · similarity 0.87)
- Joint Distribution Matters: Deep Brownian Distance Covariance for Few-Shot Classification (CVPR 2022 · similarity 0.87)
- Understanding the Covariance Structure of Convolutional Filters (ICLR 2023 · similarity 0.86)
- Discriminating image representations with principal distortions (ICLR 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.
