JackLongKing/TensorRT-CenterNet

tensorrt5 , centernet , centerface

TensorRT-CenterNet

demo (GT 1070)

• 
• 

Performance

backbone	input_size	GPU	mode	inference Time
mobilenetv2	512x512	gtx 1070	float32	3.798ms
mobilenetv2	512x512	jetson TX2	float16	22ms

Enviroments

1. gtx 1070

ubuntu 1604
TensorRT 5.0

2. jetson TX2

jetpack 4.2

Models

1. Convert CenterNet model to onnx (deform conv is not support)
2. Use netron to observe whether the output of the converted onnx model is (hm, reg, wh)

Example

git clone https://github.com/CaoWGG/TensorRT-CenterNet.git
cd TensorRT-CenterNet
mkdir build
cd build && cmake .. && make
cd ..
##cthelmet
./buildEngine -i model/ctdet_helmet.onnx -o model/ctdet_helmet.engine
./runDet -e model/ctdet_helmet.engine -i test.jpg -c test.h264
##centerface
./buildEngine -i model/centerface.onnx -o model/centerface.engine
./runDet -e model/centerface.engine -i test.jpg -c test.h264
## a simple python demo (need cv2 numpy)
python3 run.py

Related projects

• TensorRT-Yolov3
• CenterNet
• centerface
• netron
• cpp-optparse