<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>DeepStream | Mutte Ur Rehman · Independent AI Engineer</title><link>/tag/deepstream/</link><atom:link href="/tag/deepstream/index.xml" rel="self" type="application/rss+xml"/><description>DeepStream</description><generator>Wowchemy (https://wowchemy.com)</generator><language>en-us</language><lastBuildDate>Fri, 01 Mar 2024 00:00:00 +0000</lastBuildDate><image><url>/media/icon_hu1df5b22a25c153c0bc30315164a416d3_6811_512x512_fill_lanczos_center_3.png</url><title>DeepStream</title><link>/tag/deepstream/</link></image><item><title>Multi-Camera Real-Time CV Pipeline</title><link>/project/multi-camera-cv/</link><pubDate>Fri, 01 Mar 2024 00:00:00 +0000</pubDate><guid>/project/multi-camera-cv/</guid><description>&lt;p>Real-time multi-camera system in C++17 using NVIDIA DeepStream SDK and GStreamer, processing 4 or more concurrent video streams at 30+ FPS on NVIDIA Jetson edge devices. PyTorch to ONNX to TensorRT (FP16 and INT8) model conversion pipeline delivered 3x inference speedup, with OpenVINO fallback for Intel hardware. CI/CD with GitHub Actions, Docker deployment, unit tests with Google Test.&lt;/p></description></item></channel></rss>