Methodology for Distributed CNN Inference on IoT Edge Devices
- Project
- 16018 COMPACT
- Type
- New standard
- Description
- First approach for full distributed inference on IoT Edge nodes
- Pools memory resources of all devices in network
- Contact
- Daniel Mueller-Gritschneder
- daniel.mueller@tum.de
- Technical features
Input(s):
- IoT Nodes
- CNN
Main feature(s):
- Code Generation of Distributed CNN Inference Software
- Optimization
Output(s):
- CNN Inference Software
- Integration constraints
- Based on larger library, only larger Edge Devices supported
- Experimental, industrial adaption required
- Targeted customer(s)
Providers of IoT Sensor Device Networks that want to integrate AI.
- Conditions for reuse
- Methodology published
- Code not open source
- Confidentiality
- Public
- Publication date
- 08-12-2020
- Involved partners
- Technical University of Munich (DEU)