Skip to content

Latest commit

 

History

History
71 lines (62 loc) · 3.41 KB

README.md

File metadata and controls

71 lines (62 loc) · 3.41 KB
Raw

1. get onnx

download directly at weiyun or google driver or export onnx:

# 🔥 yolov8 offical repo: https://github.com/ultralytics/ultralytics
# 🔥 yolov8 quickstart: https://docs.ultralytics.com/quickstart/
# 🚀TensorRT-Alpha will be updated synchronously as soon as possible!

# install yolov8
conda create -n yolov8 python==3.8 -y # for Linux
# conda create -n yolov8 python=3.9 -y # for Windows10
conda activate yolov8
pip install ultralytics==8.0.200
pip install onnx==1.12.0

# download offical weights(".pt" file)
https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8n-pose.pt
https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8s-pose.pt
https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8m-pose.pt
https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8l-pose.pt
https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8x-pose.pt

export onnx:

yolo mode=export model=yolov8n-pose.pt format=onnx dynamic=True opset=12
yolo mode=export model=yolov8s-pose.pt format=onnx dynamic=True opset=12    
yolo mode=export model=yolov8m-pose.pt format=onnx dynamic=True opset=12    
yolo mode=export model=yolov8l-pose.pt format=onnx dynamic=True opset=12    
yolo mode=export model=yolov8x-pose.pt format=onnx dynamic=True opset=12

2.edit and save onnx

# note: If you have obtained onnx by downloading, this step can be ignored
ignore

3.compile onnx

# put your onnx file in this path:tensorrt-alpha/data/yolov8-pose
cd tensorrt-alpha/data/yolov8-pose
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/home/feiyull/TensorRT-8.4.2.4/lib
../../../../TensorRT-8.4.2.4/bin/trtexec   --onnx=yolov8n-pose.onnx  --saveEngine=yolov8n-pose.trt  --buildOnly --minShapes=images:1x3x640x640 --optShapes=images:2x3x640x640 --maxShapes=images:4x3x640x640
../../../../TensorRT-8.4.2.4/bin/trtexec   --onnx=yolov8s-pose.onnx  --saveEngine=yolov8s-pose.trt  --buildOnly --minShapes=images:1x3x640x640 --optShapes=images:2x3x640x640 --maxShapes=images:4x3x640x640
../../../../TensorRT-8.4.2.4/bin/trtexec   --onnx=yolov8m-pose.onnx  --saveEngine=yolov8m-pose.trt  --buildOnly --minShapes=images:1x3x640x640 --optShapes=images:2x3x640x640 --maxShapes=images:4x3x640x640
../../../../TensorRT-8.4.2.4/bin/trtexec   --onnx=yolov8l-pose.onnx  --saveEngine=yolov8l-pose.trt  --buildOnly --minShapes=images:1x3x640x640 --optShapes=images:2x3x640x640 --maxShapes=images:4x3x640x640
../../../../TensorRT-8.4.2.4/bin/trtexec   --onnx=yolov8x-pose.onnx  --saveEngine=yolov8x-pose.trt  --buildOnly --minShapes=images:1x3x640x640 --optShapes=images:2x3x640x640 --maxShapes=images:4x3x640x640

4.run

git clone https://github.com/FeiYull/tensorrt-alpha
cd tensorrt-alpha/yolov8-pose
mkdir build
cd build
cmake ..
make -j10
# note: the dstImage will be saved in tensorrt-alpha/yolov8-pose/build by default

## 640
# infer image
./app_yolov8_pose  --model=../../data/yolov8/yolov8n-pose.trt --size=640 --batch_size=1  --img=../../data/6406407.jpg   --show --savePath=../

# infer video
./app_yolov8_pose  --model=../../data/yolov8/yolov8n-pose.trt     --size=640 --batch_size=2  --video=../../data/people.mp4  --show 

# infer camera
./app_yolov8_pose  --model=../../data/yolov8/yolov8n-pose.trt     --size=640 --batch_size=2  --cam_id=0  --show

5. appendix

ignore