Workshop1: AI for Everyone
summary
Huawei outlines an on‑device AI stack that compiles neural nets into offline compiled models for the Kirin 970 NPU, claiming up to 3× faster Prisma-style style transfer than iPhone X cloud and sub‑1s targets; TikTok-style segmentation shows much tighter edges on an S9. Current runtime supports about 90 operators supported (plans for ~150), model manager lets apps download/replace models, and ResNet‑50 runs ~2.5 FPS on a P20 Pro vs ~8–9 on iPhone X.
AI-written summary · about
chapters (13)
- 0:00Introduction to AI for Everyone
- 3:15Mobile AI trends and challenges
- 7:39Huawei AI mobile computing platform
- 14:43Technology behind HiAI Foundation
- 20:34Benefits of on-device AI
- 23:22Q&A session part 1
- 29:29Introduction to HiAI Engine
- 36:16Challenges of AI development
- 41:12HiAI Engine features and APIs
- 55:30WPS Office AI integration demo
- 1:00:58HiAI Engine developer resources
- 1:02:13Live coding with HiAI APIs
- 1:14:02Q&A session part 2
related talks
29mMake your app smarter
24mMake your app smarter
31mBringing AI on-Device: From Cloud to Edge
36mHuaweu HiAI Unlock The Future
54mHarnessing On-Device AI to Transform the Smartphone Experience
32mAI for Mobile and IoT Devices: TensorFlow Lite
46mPowering On-device AI Capabilities with Samsung Neural SDK
20mProductionizing On-device AI Models