Re-enables INT8 inference (using experimental/filesystem for C++14 compatibility)
Adds benchmarks for yolov8m and yolov8x. INT8 precision will follow.
Since you are not using a lot of the C++17 features, I would suggest you to retro-back to C++14 also in main (for better system compatibility with users), as you want.
Still working on dynamic tensor inputs on the old TensorRT API.
Hi,
This PR adds the following:
experimental/filesystem
for C++14 compatibility)yolov8m
andyolov8x
. INT8 precision will follow.Since you are not using a lot of the C++17 features, I would suggest you to retro-back to C++14 also in
main
(for better system compatibility with users), as you want.Still working on dynamic tensor inputs on the old TensorRT API.
Thanks,