Research map / Detection and segmentation / CNN features
Bi-volution: A Static and Dynamic Coupled Filter
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 |
|---|---|---|---|
| Combine static and dynamic convolution in a dual-branch operator | ResNet-50 · ImageNet validation | top-1 accuracy: 76.8 -> 79.6 | Helped |
| Use Bi-volution instead of ResNet-50 convolution | Faster R-CNN · COCO 2017 validation | box AP: 37.7 -> 40.1 | Helped |
| Use Bi-volution instead of ResNet-50 convolution | Mask R-CNN · COCO 2017 validation | mask AP: 35.1 -> 36.4 | Helped |
| Combine static and dynamic convolution | BiNet-50 · ImageNet-C corruptions | average accuracy: 47.5% -> 48.7% | Helped |
| Combine static and dynamic convolution | BiNet-50 · ImageNet-C noise corruptions | average accuracy: +3.0% | Helped |
| Add the static-convolution branch (ablation) | BiNet-50 (dc) · ImageNet classification | top-1 accuracy: 78.1 -> 78.2 | Helped |
| Add the Spatial Augmented Kernel Generation module (ablation) | BiNet-50 (dc) · ImageNet classification | top-1 accuracy: 78.1 -> 78.5 | Helped |
| Add both the static branch and SAKG (ablation) | BiNet-50 (dc) · ImageNet classification | top-1 accuracy: 78.1 -> 78.8 | Helped |
| Add the static-convolution branch (ablation) | BiNet-50 (dc-ca) · ImageNet classification | top-1 accuracy: 79.1 -> 79.3 | Helped |
| Add the Spatial Augmented Kernel Generation module (ablation) | BiNet-50 (dc-ca) · ImageNet classification | top-1 accuracy: 79.1 -> 79.3 | Helped |
| Add both the static branch and SAKG (ablation) | BiNet-50 (dc-ca) · ImageNet classification | top-1 accuracy: 79.1 -> 79.4 | Helped |
| Expand the SAKG kernel from 1×1 to 3×3 (ablation) | BiNet-50 (dc) · ImageNet classification | top-1 accuracy: 78.1 -> 78.5 | Helped |
Closest papers by meaning
- Adaptive Dynamic Filtering Network for Image Denoising (AAAI 2023 · similarity 0.88)
- KernelWarehouse: Rethinking the Design of Dynamic Convolution (ICML 2024 · similarity 0.87)
- Involution: Inverting the Inherence of Convolution for Visual Recognition (CVPR 2021 · similarity 0.87)
- Dynamic Filter Networks (NeurIPS 2016 · similarity 0.87)
- Frequency Dynamic Convolution for Dense Image Prediction (CVPR 2025 · similarity 0.85)
- SELECT: A Large-Scale Benchmark of Data Curation Strategies for Image Classification (NeurIPS 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.