Geekgineer/YOLOs-CPP-TensorRT

YOLOs-TRT is a header-only C++ library for running all YOLO models with all tasks with NVIDIA TensorRT on CUDA GPUs and Jetson. It features GPU preprocessing (letterbox/normalize/HWC→NCHW), CUDA Graph replay, FP16/INT8 support, and sub-2ms end-to-end latency with 530+fps.

语言：C++ · GitHub 星标：92 · Fork：9

在 GitHub 上查看原仓库