This repository provides demos for running optimized deep learning models on NVIDIA hardware using TensorRT. It covers multiple computer vision tasks including object detection with YOLOv3 and YOLOv4, SSD-based detection, MTCNN face detection, GoogLeNet image classification, and MODNet video matting. The demos target embedded NVIDIA Jetson platforms as well as x86_64 PCs with NVIDIA GPUs, focusing on practical inference performance optimization.