Research map / Architectures and efficiency / CNN design
Convolution of Convolution: Let Kernels Spatially Collaborate
Rongzhen Zhao, Jian Li, Zhenzhi Wu
Check 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 |
|---|---|---|---|
| add CoC | ImageNet classification backbones · ImageNet/val | accuracy: 70.01 -> 70.40; 76.25 -> 76.45; 74.03 -> 74.25; 75.82 -> 76.07; 69.59 -> 70.24; 66.34 -> 66.84 | Improved |
| combine CoC with orthogonalization | ResNet18 · ImageNet/val | accuracy: 70.01 -> 71.20 | Improved |
| combine CoC with orthogonalization | ResNet50 · ImageNet/val | accuracy: 76.25 -> 77.07 | Improved |
| add CoC | RetinaNet-r50 · COCO2017/val bbox | mAP: 36.3 -> 38.1 | Improved |
| add CoC | MaskRCNN-hr18 · COCO2017/val mask | mAP: 33.9 -> 35.6 | Improved |
| add CoC | FCN-r50d8 · VOC0712/val | mIoU: 66.97 -> 69.52 | Improved |
| add CoC | HRNet-w18 · VOC0712/val | mIoU: 72.13 -> 74.81 | Improved |
| add CoC while training from scratch | FCN-r50d8 · VOC0712/val | mIoU: 23.53 -> 34.25 | Improved |
| add CoC while training from scratch | HRNet-w18 · VOC0712/val | mIoU: 35.23 -> 41.39 | Improved |
| increase spatial association size beyond sa=2 (ablation) | CoC · Tiny-ImageNet | validation accuracy | No clear effect |
| increase super-kernel expansion too far (ablation) | CoC · Tiny-ImageNet | validation accuracy | Hurt |
| use more than one super layer (ablation) | CoC · Tiny-ImageNet | validation accuracy | Hurt |
Closest papers by meaning
- Involution: Inverting the Inherence of Convolution for Visual Recognition (CVPR 2021 · similarity 0.88)
- Non-Linear Convolution Filters for CNN-Based Learning (ICCV 2017 · similarity 0.87)
- End-to-End Kernel Learning with Supervised Convolutional Kernel Networks (NeurIPS 2016 · similarity 0.87)
- Approximation and Learning with Deep Convolutional Models: a Kernel Perspective (ICLR 2022 · similarity 0.87)
- KernelDNA: Cross-Layer Kernel Sharing via Decoupled Neural Adapters (NeurIPS 2026 · similarity 0.87)
- KernelWarehouse: Rethinking the Design of Dynamic Convolution (ICML 2024 · 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.
