Research map / Architectures and efficiency / CNN design
CNN Filter DB: An Empirical Investigation of Trained Convolutional Filters
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 |
|---|---|---|---|
| Increase model capacity beyond what the simple low-resolution dataset requires | CNN classifiers trained on CIFAR-10 and CIFAR-100 · CIFAR-10 and CIFAR-100 | filter sparsity and entropy | Hurt |
| Use robust training | ImageNet1k classifiers · ImageNet1k | filter diversity and sparsity | Helped |
| Compare trained filters across tasks | CNN models across tasks · Multiple vision tasks | filter distribution drift | No clear effect |
| Compare models trained on different visual categories | CNN models across visual categories · Multiple visual categories | filter distribution drift | No clear effect |
| Use GAN discriminators rather than other task models | CNN filter distributions · GAN discriminators versus other tasks | transferability of filter distributions | Hurt |
| Compare classification filters by convolution depth | Classification models · Convolution depth | model-to-model filter drift | Mixed |
| Compare object and face detection filters by convolution depth | Object and face detection models · Convolution depth | model-to-model filter drift | Mixed |
| Retrain ResNet-9 with different random seeds | ResNet-9 · CIFAR-10 | filter distribution drift D: D < 5.3\cdot 10^{-4} | No clear effect |
| Retrain MobileNetv2 with different random seeds | MobileNetv2 · CIFAR-10 | filter distribution drift D: D < 2.6\cdot 10^{-2} | Mixed |
| Train with fp16 instead of fp32 | Low-resolution CNN models · CIFAR-10 | filter distribution drift D | No clear effect |
| Exclude filters from layers classified as degenerated (ablation) | CNN Filter DB shift analysis · Tasks and visual categories | correlation between degeneration ratio and mean shift | No clear effect |
Closest papers by meaning
- RePr: Improved Training of Convolutional Filters (CVPR 2019 · similarity 0.91)
- Approximation and Learning with Deep Convolutional Models: a Kernel Perspective (ICLR 2022 · similarity 0.91)
- Deep Edge Filter: Return of the Human-Crafted Layer in Deep Learning (NeurIPS 2025 · similarity 0.90)
- Unveiling the Unseen: Identifiable Clusters in Trained Depthwise Convolutional Kernels (ICLR 2024 · similarity 0.89)
- KernelWarehouse: Rethinking the Design of Dynamic Convolution (ICML 2024 · similarity 0.89)
- The Master Key Filters Hypothesis: Deep Filters Are General (AAAI 2025 · similarity 0.88)
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.