Research map / Architectures and efficiency / CNN design
Constructing Fast Network through Deconstruction of Convolution
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 heuristic grouped shifts with learned real-valued shifts (ablation) | AS-ResNet-w32 · ImageNet | Top-1 accuracy: 59.8 -> 64.1 | Helped |
| use randomly sampled integer shifts instead of heuristic grouped shifts (ablation) | AS-ResNet-w32 · ImageNet | Top-1 accuracy: 59.8 -> 60.1 (+0.3) | Helped |
| use randomly sampled real-valued shifts instead of heuristic grouped shifts (ablation) | AS-ResNet-w32 · ImageNet | Top-1 accuracy: 59.8 -> 61.9 (+2.1) | Helped |
| learn the real-valued shifts rather than keeping them fixed (ablation) | AS-ResNet-w32 · ImageNet | Top-1 accuracy: 61.9 -> 64.1 | Helped |
| replace depthwise convolution block with an active shift layer block | CIFAR-10 network · CIFAR-10 | accuracy: 94.16 -> 94.5 | Helped |
| replace depthwise convolution block with an active shift layer block | CIFAR-10 network · CIFAR-10 | CPU inference time: 16 ms -> 10.6 ms | Helped |
| widen the base width rather than increasing expansion rate | ASNet · CIFAR-10 | accuracy: 92.93 -> 93.52 | Helped |
| widen the base width rather than increasing expansion rate | ASNet · CIFAR-100 | accuracy: 71.83 -> 73.07 | Helped |
| use AS-ResNet-w68 instead of MobileNetV2 | ImageNet classifier · ImageNet | CPU inference time: 54.7 -> 47.9 | Helped |
| use AS-ResNet-w68 instead of MobileNetV2 | ImageNet classifier · ImageNet | Top-1 accuracy: 71.8 -> 72.2 | Helped |
Closest papers by meaning
- All You Need Is a Few Shifts: Designing Efficient Convolutional Neural Networks for Image Classification (CVPR 2019 · similarity 0.89)
- Shift: A Zero FLOP, Zero Parameter Alternative to Spatial Convolutions (CVPR 2018 · similarity 0.88)
- FreeNet: Liberating Depth-Wise Separable Operations for Building Faster Mobile Vision Architectures (AAAI 2025 · similarity 0.84)
- AsCAN: Asymmetric Convolution-Attention Networks for Efficient Recognition and Generation (NeurIPS 2024 · similarity 0.84)
- LSNet: See Large, Focus Small (CVPR 2025 · similarity 0.83)
- DECO: Unleashing the Potential of ConvNets for Query-based Detection and Segmentation (ICLR 2025 · similarity 0.83)
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.