Research map / Architectures and efficiency / RNNs and normalization
EtinyNet: Extremely Tiny Network for TinyML
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 |
|---|---|---|---|
| Use EtinyNet-1.0 instead of MobileNeXt-0.35 | ImageNet-1000 classification · ImageNet-1000 | Top-1 accuracy: 64.7 -> 65.5 | Helped |
| Use EtinyNet-0.75 (4-bit) with ASQ (ablation) | EtinyNet-0.75 · ImageNet-1000 | Top-1 accuracy: 45.3% -> 57.0% | Helped |
| Use EtinyNet-1.0 (4-bit) with ASQ (ablation) | EtinyNet-1.0 · ImageNet-1000 | Top-1 accuracy: 55.1% -> 60.7% | Helped |
| Replace the proposed LB/DLB blocks with inverted bottleneck blocks (ablation) | EtinyNet · ImageNet-1000 | Top-1 accuracy: 2.4% | Hurt |
| Add ReLU between depthwise and pointwise convolution (ablation) | EtinyNet · ImageNet-1000 | Top-1 accuracy: 0.8% | Hurt |
| Deploy EtinyNet-1.0 instead of MCUNet | STM32F412 ImageNet classification · ImageNet-1000 on STM32F412 | Top-1 accuracy: 62.2% -> 64.7% | Helped |
| Deploy EtinyNet-1.0 instead of MCUNet | STM32F746 ImageNet classification · ImageNet-1000 on STM32F746 | Top-1 accuracy: 63.5% -> 65.8% | Helped |
| Use EtinyNet-SSD instead of MCUNet | STM32H743 object detection · Pascal VOC | mAP: 51.4% -> 56.4% | Helped |
| Deploy the compact EtinyNet-0.75 (4-bit) FPGA design instead of AngelEye | FPGA ImageNet inference · FPGA ImageNet inference | power: 3.5 -> 0.62 | Helped |
Closest papers by meaning
- AtomNet: Designing Tiny Models from Operators Under Extreme MCU Constraints (AAAI 2025 · similarity 0.74)
- An Empirical Study of Distributed Deep Learning Training on Edge (Student Abstract) (AAAI 2024 · similarity 0.74)
- TinyTTA: Efficient Test-time Adaptation via Early-exit Ensembles on Edge Devices (NeurIPS 2024 · similarity 0.73)
- TinyTrain: Resource-Aware Task-Adaptive Sparse Training of DNNs at the Data-Scarce Edge (ICML 2024 · similarity 0.72)
- DEX: Data Channel Extension for Efficient CNN Inference on Tiny AI Accelerators (NeurIPS 2024 · similarity 0.72)
- Demystifying Cost-Efficiency in LLM Serving over Heterogeneous GPUs (ICML 2025 · similarity 0.71)
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.