Computer Vision at the Edge with OpenVINO
summary
OpenVINO converts trained Caffe/TensorFlow/MXNet models to an intermediate IR (XML + .bin) with a model optimizer, then runs inference via a plugin-based engine on CPU, integrated GPU or Myriad VPU without recoding. Demo pipelines run face detection and a 3-stage vehicle → plate → OCR sequence on Ubuntu 16.04, hitting 40–45 FPS on i7 and about 15–20 FPS on low‑end boards; Movidius requires FP16 on Movidius models and exposes per-layer performance_count for profiling.
AI-written summary · about
chapters (5)
- 0:00Intro
- 1:09Accelerate Computer Vision with OpenVINO Toolkit
- 2:17Benefits of the OpenVINO toolkit
- 3:55OpenVINO toolkit Technical Specifications
- 6:20Improve Performance with Model Optimizer
related talks
23mOptimize your network inference time with OpenVINO
26mDeep Learning in OpenCV
42mAI Everywhere– accelerate your development from PC to cloud
13mAccelerating AI algorithms
14mComputer vision on edge devices
59mML and the IoT: Living on the Edge
50mAI acceleration at the edge
16mRVEdge-Vision: A Fully Open,Ultra-Efficient On-Device AI Platform for Smart Eyewear