Vision AI at the Edge for Industrial Inspection
summary
Neruala inspection software runs on QCS6490/QCM6490 boards (Windows 11 IoT Enterprise supported) and uses the Qualcomm Neural Processing SDK plus the AI Model Efficiency Toolkit to compress and quantize a 10-class industrial anomaly detector into a INT8 quantized model (DLC runtime). Accuracy stays within a few percent of FP32 while throughput and power improve, delivering sub-millisecond latency on edge hardware with triple ISP, integrated 5G and hardware root of trust.
AI-written summary · about
chapters (11)
- 0:00Introduction to edge AI
- 1:37Connected intelligent edge
- 3:23Qualcomm core pillars
- 5:23Unified chipset solutions
- 6:42Deployment architectures
- 8:43Chipset technology highlights
- 12:14Software stack and tools
- 15:06Anomaly detection case study
- 17:34Development workflow
- 20:46Performance results
- 23:16Developer resources
related talks
08mUnlocking AI at Scale for PyTorch Developers
16mEnabling AI on Low Power Endpoint Devices -QuickLogic & SiFive Freedom Aware
50mAI acceleration at the edge
14mEmbedded Executive LIVE
28mML on the Edge for Industry 4.0 with Arm Ethos N-78 Neural Processing Unit
54mHarnessing On-Device AI to Transform the Smartphone Experience
04mEmbedded Executive LIVE
16mRVEdge-Vision: A Fully Open,Ultra-Efficient On-Device AI Platform for Smart Eyewear
