Research map / Architectures and efficiency / CNN design
ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices
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 |
|---|---|---|---|
| Add channel shuffle to grouped pointwise convolutions (ablation) | ShuffleNet · ImageNet 2012 validation | classification error (%): 34.5 -> 32.6; 37.6 -> 32.4; 45.7 -> 43.2; 48.1 -> 42.3; 56.3 -> 55.0; 56.5 -> 52.7 | Helped |
| Use pointwise group convolution rather than dense pointwise convolution (ablation) | ShuffleNet · ImageNet 2012 validation | classification error (%): 33.6 -> 32.4; 45.1 -> 41.6; 57.1 -> 52.7 | Helped |
| Increase pointwise group count to 8 rather than 4 (ablation) | ShuffleNet 0.5× · ImageNet 2012 validation | classification error (%): 41.6 -> 42.3 | Hurt |
| Increase pointwise group count to 8 rather than 4 (ablation) | ShuffleNet 0.25× · ImageNet 2012 validation | classification error (%): 54.2 -> 52.7 | Helped |
| Add SE blocks (ablation) | ShuffleNet 2× (g=3) · ImageNet 2012 validation | classification error (%): 26.3 -> 24.7 | Helped |
| Add SE blocks (ablation) | ShuffleNet 2× (g=3) · Snapdragon 820 mobile inference | inference time: 25 ~ 40% slower | Hurt |
| Use ShuffleNet instead of MobileNet at similar compute budgets | ImageNet classification models · ImageNet 2012 validation | classification error (%): 29.4 -> 26.3; 36.3 -> 32.4; 49.4 -> 41.6 | Helped |
| Use ShuffleNet 2× instead of MobileNet 1.0× | Faster R-CNN detector · MS COCO minival, 600× image | mAP [.5,.95]: 19.8% -> 25.0% | Helped |
Closest papers by meaning
- MobileODE: An Extra Lightweight Network (NeurIPS 2025 · similarity 0.88)
- MnasNet: Platform-Aware Neural Architecture Search for Mobile (CVPR 2019 · similarity 0.88)
- MobileOne: An Improved One Millisecond Mobile Backbone (CVPR 2023 · similarity 0.87)
- IFORMER: INTEGRATING CONVNET AND TRANSFORMER FOR MOBILE APPLICATION (ICLR 2025 · similarity 0.85)
- FreeNet: Liberating Depth-Wise Separable Operations for Building Faster Mobile Vision Architectures (AAAI 2025 · similarity 0.85)
- RepViT: Revisiting Mobile CNN From ViT Perspective (CVPR 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.