Optimize your network inference time with OpenVINO
summary
Converts trained TensorFlow/Keras/PyTorch (via ONNX) models with the Model Optimizer into IR files (XML + BIN), applies optional post‑training quantization, and runs inference with OpenVINO Runtime on CPU, integrated GPU, or VPU. Demonstrated pipeline uses omz downloader/converter, benchmark_app and notebooks, with measured throughput like 55 fps on CPU and 65 fps on iGPU for common nets (ResNet50, YOLOv3‑tiny, DeepLabv3), supports FP32/FP16/INT8 precisions and devices including Movidius NCS2 VPU and shows the end‑to‑end conversion to IR XML+BIN.
AI-written summary · about
chapters (9)
- 0:00Start
- 3:41Installation Methods
- 4:56Model Optimizer
- 5:58OpenVINO Runtime
- 6:47Supported Devices
- 8:22AUTO Device
- 17:24OpenVINO Notebooks
- 18:16Intel® DevCloud for the Edge
- 19:04Platform Configurations for Performance Benchmarks
related talks
20mComputer Vision at the Edge with OpenVINO
29mEfficient AI with Python: running AI models on CPUs
11mLightning Talk: Accelerating PyTorch Performance with OpenVINO
18mDeploying ML solutions with low latency in Python
42mAI Everywhere– accelerate your development from PC to cloud
30mScalable ML acceleration with ONNX Runtime
24mTools & Techniques for a cost optimized Analytics & AI pipeline
23mThe Race to the Bottom: Low Latency in the age of the Transformer