TensorRT-CenterNet
demo (GT 1070)
Performance
| backbone |
input_size |
GPU |
mode |
inference Time |
| mobilenetv2 |
512x512 |
gtx 1070 |
float32 |
3.798ms |
| mobilenetv2 |
512x512 |
gtx 1070 |
int8 |
1.75ms |
| mobilenetv2 |
512x512 |
jetson TX2 |
float16 |
22ms |
Enviroments
- gtx 1070
- jetson TX2
Models
- Convert CenterNet model to onnx (deform conv is not support)
- Use netron to observe whether the output of the converted onnx model is (hm, reg, wh)
- for centerface
# a simple way to change input shape of (model.onnx)
import onnx
input_size = (512,512)
model = onnx.load_model("centerface.onnx")
d = model.graph.input[0].type.tensor_type.shape.dim
rate = (input_size[0]//d[2].dim_value,input_size[1]//d[3].dim_value)
d[0].dim_value = 1
d[2].dim_value *= rate[0]
d[3].dim_value *= rate[1]
for output in model.graph.output:
d = output.type.tensor_type.shape.dim
d[0].dim_value = 1
d[2].dim_value *= rate[0]
d[3].dim_value *= rate[1]
onnx.save_model(model,"centerface_changed_shape.onnx")Example
git clone https://github.com/CaoWGG/TensorRT-CenterNet.git
cd TensorRT-CenterNet
mkdir build
cd build && cmake .. && make
cd ..
##cthelmet | config include/ctdetConfig.h
## flaot32
./buildEngine -i model/ctdet_helmet.onnx -o model/ctdet_helmet.engine -m 0
./runDet -e model/ctdet_helmet.engine -i test.jpg -c test.h264
## int8
./buildEngine -i model/ctdet_helmet.onnx -o model/ctdet_helmet.engine -m 2 -c calib_img_list.txt
./runDet -e model/ctdet_helmet.engine -i test.jpg -c test.h264
##centerface | config include/ctdetConfig.h
./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
