Keynote: WEdge Project: Edge AI enabled by Wasm
Talkat WasmCon 2023 · on Thursday, 2023-09-07· at 09:50· Grand Ballrooms A-D, Hyatt Regency Bellevue, 900 Bellevue Way NE·
summary
Midokura builds WEdge as an edge AI platform adopted by Sony’s AITRIOS and already in commercial use, claiming Wasm gives memory isolation, plugin-based deploys and safer over-the-air updates. The cloud does AOT compilation in cloud per target device while device agents run a Whammer runtime on Nuttx or Linux, exposing OpenCV, blob storage and wazzy sensor API for CMOS sensors and DSPs like the IMX500; current limits include piedoism limited Python support while upstream Whammer contributions and open sourcing are planned.
AI-written summary · about
chapters (10)
- 0:00WEdge Project intro
- 1:08Sony acquisition
- 2:58Motivation for WEdge
- 4:27IoT development challenges
- 6:14WEdge goals and vision
- 7:56WEdge Cloud architecture
- 8:37WEdge Agent runtime
- 9:21WEdge services and APIs
- 10:46Polyglot Wasm challenges
- 12:10Future plans and community
related talks
07mKeynote: Modernizing the Development of AI IoT Devices with WebAssembly-based...- 07mKeynote: Modernizing the Development of AI IoT Devices with WebAssembly-based...
36mWebAssembly-based Edge AI Sensing Pipeline Orches...
38mStreamlining AI Solutions Development: S...
06mProject Lightning Talk: What's New in WasmEdge 0.15.0?
42mSecure and Efficient Sensing Applications with Wasm: Sony...
32mWebAssembly Based AI as a Service on the Edge with Kubernetes
10mLightning Talk: Scaling AI With Wasm and Edge Computing