Research map / Architectures and efficiency / CNN design
Revisiting Sparse Convolutional Model for Visual Recognition
Read the paperCheck 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 |
|---|---|---|---|
| replace first conv layer with convolutional sparse coding layer (SDNet-18) | ResNet-18 · CIFAR-100 | accuracy: 77.82% -> 78.31% | Helped |
| replace first conv layer with CSC-layer | ResNet-18 · CIFAR-10 | accuracy: 95.54% -> 95.20% | No clear effect |
| replace first conv layer with CSC-layer | ResNet-18 · ImageNet | top-1 accuracy: 68.98% -> 69.47% | Helped |
| adapt lambda at test time via Algorithm 1 (ablation) | SDNet-18 · CIFAR-10-C Gaussian level-4 | accuracy: 27.92% -> 48.84% | Helped |
| tune lambda against PGD perturbation (ablation) | SDNet-18 · CIFAR-10 PGD Linf | robust accuracy: 0.11% -> 35.18% | Helped |
| increase FISTA iterations from 2 to 8 (ablation) | SDNet-18 · ImageNet-C Impulse | accuracy: 22.01% -> 25.69% | Helped |
| increase FISTA iterations from 2 to 8 (ablation) | SDNet-18 · ImageNet | natural accuracy: 69.47% -> 69.79% | No clear effect |
| replace all conv layers with CSC-layers (ablation) | SDNet-18 · ImageNet | top-1 accuracy: 69.47% -> 69.37% | No clear effect |
| replace all conv layers with CSC-layers (ablation) | SDNet-34 · ImageNet | top-1 accuracy: 72.67% -> 72.54% | No clear effect |
Closest papers by meaning
- SparseFormer: Sparse Visual Recognition via Limited Latent Tokens (ICLR 2024 · similarity 0.90)
- A General Efficient Hyperparameter-Free Algorithm for Convolutional Sparse Learning (AAAI 2017 · similarity 0.89)
- Designing BERT for Convolutional Networks: Sparse and Hierarchical Masked Modeling (ICLR 2023 · similarity 0.89)
- From Flat to Hierarchical: Extracting Sparse Representations with Matching Pursuit (NeurIPS 2025 · similarity 0.88)
- Dynamic Sparse Training versus Dense Training: The Unexpected Winner in Image Corruption Robustness (ICLR 2025 · similarity 0.87)
- How Deep Networks Learn Sparse and Hierarchical Data: the Sparse Random Hierarchy Model (ICML 2024 · similarity 0.87)
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.