#yolov8

Achieving 165 FPS with Zero-Host-Copy Inference Pipelines

A new C++/CUDA inference pipeline architecture achieves 165 FPS by eliminating CPU-side data copying and dynamic memory allocation during the hot execution phase.