YOLOv8 using TensorRT accelerate !

1. Install TensorRT follow TensorRT offical website

2. Install python requirement.

   pip install -r requirement.txt

3. (optional) Install ultralytics YOLOv8 package for TensorRT API building.

   pip install -i https://test.pypi.org/simple/ ultralytics

   You can download pretrained pytorch model by:

   wget https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8n.pt
   wget https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8s.pt
   wget https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8m.pt
   wget https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8l.pt
   wget https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8x.pt

You can dowload the onnx model which are exported by YOLOv8 package and modified by me .

YOLOv8-n

YOLOv8-s

YOLOv8-m

YOLOv8-l

YOLOv8-x

You can export TensorRT engine from ONNX by build.py .

Usage:

python build.py \
--weights yolov8s_nms.onnx \
--iou-thres 0.65 \
--conf-thres 0.25 \
--topk 100 \
--fp16  \
--device cuda:0

• --weights : The ONNX model you download.

• --iou-thres : IOU threshold for NMS plugin.

• --conf-thres : Confidence threshold for NMS plugin.

• --topk : Max number of detection bboxes.

• --fp16 : Whether to export half-precision engine.

• --device : The CUDA deivce you export engine .

You can modify iou-thres conf-thres topk by yourself.

You can export TensorRT engine by trtexec tools.

Usage:

/usr/src/tensorrt/bin/trtexec --onnx=yolov8s_nms.onnx --saveEngine=yolov8s_nms.engine --fp16

If you installed TensorRT by a debian package, then the installation path of trtexec is /usr/src/tensorrt/bin/trtexec

If you installed TensorRT by a tar package, then the installation path of trtexec is under the bin folder in the path you decompressed

When you want to build engine by api. You should generate the pickle weights parameters first.

python gen_pkl.py -w yolov8s.pt -o yolov8s.pkl

You will get a yolov8s.pkl which contain the operators' parameters. And you can rebuild yolov8s model in TensorRT api.

python build.py \
--weights yolov8s.pkl \
--iou-thres 0.65 \
--conf-thres 0.25 \
--topk 100 \
--fp16  \
--input-shape 1 3 640 640 \
--device cuda:0

Notice !!! Now we only support static input shape model build by TensorRT api. You'd best give the legalinput-shape.

You can infer images with the engine by infer.py .

Usage:

python3 infer.py --engine yolov8s_nms.engine --imgs data --show --out-dir outputs --device cuda:0

• --engine : The Engine you export.
• --imgs : The images path you want to detect.
• --show : Whether to show detection results.
• --out-dir : Where to save detection results images. It will not work when use --show flag.
• --device : The CUDA deivce you use.
• --profile : Profile the TensorRT engine.

You can infer with c++ in csrc/end2end .

Build:

Please set you own librarys in CMakeLists.txt and modify you own config in yolov8.hpp such as classes names and colors .

export root=${PWD}
cd src/end2end
mkdir build
cmake ..
make
mv yolov8 ${root}
cd ${root}

Usage:

# infer image
./yolov8 yolov8s_nms.engine data/bus.jpg
# infer images
./yolov8 yolov8s_nms.engine data
# infer video
./yolov8 yolov8s_nms.engine data/test.mp4 # the video path

If you want to profile the TensorRT engine:

Usage:

python3 infer.py --engine yolov8s_nms.engine --profile