Searching Neural Architectures for Sensor Nodes on IoT Gateways
A new paper introduces a method for designing neural networks directly on IoT gateways, enabling edge-based machine learning for privacy-sensitive applications. This approach allows for the creation of custom, hardware-friendly neural networks without sending data to the cloud, which is particularly beneficial for sectors like Healthcare IoT and Industrial IoT. Experiments on the Visual Wake Words dataset demonstrated that this method can achieve state-of-the-art results within a 10-hour search period on a Raspberry Pi Zero 2. AI
IMPACT Enables privacy-preserving AI applications in sensitive IoT environments like healthcare and industry.
- Hugging Face
- Internet of Things
- Raspberry Pi Zero 2 W
- DagsHub
- alphaXiv
- CORE Recommender
- ScienceCast
- industrial internet of things
- Gotit.pub
- CatalyzeX Code Finder for Papers
- IArxiv Recommender
- Andrea Mattia Garavagno
- Visual Wake Words dataset
- Healthcare Internet of Things: Security Threats, Challenges and Future Research Directions