Research map / Architectures and efficiency / CNN design
DeepRebirth: Accelerating Deep Neural Network Execution on 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 |
|---|---|---|---|
| Apply streamline and branch slimming | GoogLeNet · Samsung Galaxy S5 | forwarding time: 651.4 ms -> 210.6 ms | Helped |
| Apply streamline and branch slimming | GoogLeNet · Samsung Galaxy S5 | runtime memory: 33.2 MB -> 13.2 MB | Helped |
| Add Tucker decomposition after slimming | GoogLeNet-Slim · Samsung Galaxy S5 | forwarding time: 210.6 ms -> 106.3 ms | Helped |
| Apply Tucker decomposition to the original network without slimming | GoogLeNet · Samsung Galaxy S5 | forwarding time: 651.4 ms -> 614.9 ms | No clear effect |
| Apply Tucker decomposition to the slim model | GoogLeNet-Slim · ImageNet ILSVRC-2012 validation | Top-5 accuracy: 88.43% -> 86.54% | Hurt |
| Slim conv1 and conv2 with their following normalization layers | AlexNet · Samsung Galaxy S5 | forwarding time: 445 ms -> 274 ms | Helped |
| Slim conv1 and conv2 with their following normalization layers | AlexNet · ImageNet | Top-5 accuracy: 80.03% -> 79.57% | Hurt |
| Slim conv1 and res2a | ResNet-50 · Samsung Galaxy S5, processed layers conv1 and res2a | latency: 189 ms -> 104 ms | Helped |
| Slim conv1 and res2a | ResNet-50 · ImageNet | Top-5 accuracy: 92.36% -> 91.88% | Hurt |
| Slim the GoogLeNet inception stages cumulatively (ablation) | GoogLeNet · ImageNet ILSVRC-2012 validation | Top-5 accuracy: 88.89% -> 88.43% | Mixed |
| Reduce AlexNet conv1 and conv2 output paths by merging norm1 and norm2 (ablation) | AlexNet · Samsung Galaxy S5 | energy cost: 688 mJ -> 458 mJ | Helped |
| Slim ResNet-50 conv1 and res2a stages cumulatively (ablation) | ResNet-50 · ResNet (conv1-res2a) | runtime memory: 2505 MB -> 1133 MB | Helped |
Closest papers by meaning
- RT3D: Achieving Real-Time Execution of 3D Convolutional Neural Networks on Mobile Devices (AAAI 2021 · similarity 0.87)
- Fast Sparse ConvNets (CVPR 2020 · similarity 0.85)
- MobileOne: An Improved One Millisecond Mobile Backbone (CVPR 2023 · similarity 0.85)
- FreeNet: Liberating Depth-Wise Separable Operations for Building Faster Mobile Vision Architectures (AAAI 2025 · similarity 0.83)
- TinyTrain: Resource-Aware Task-Adaptive Sparse Training of DNNs at the Data-Scarce Edge (ICML 2024 · similarity 0.83)
- SLiM: One-shot Quantization and Sparsity with Low-rank Approximation for LLM Weight Compression (ICML 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.