tensorRT for Yolov3
Ubuntu 16.04
TensorRT 4.0.1.6
CUDA 9.0
Download the caffe model converted by official model:
#build source code
git submodule update --init --recursive
mkdir build
cd build && cmake .. && make && make install
cd ..
#for mobilenet-yolov3-608
> sh demo.sh
Model | GPU | Mode | Inference Time |
---|---|---|---|
mobilenet-yolov3-608 | GTX 1080 | float32 | 12.655ms |
mobilenet-yolov3-608 | Jetson nano | float32 | 247.655ms |
mobilenet-yolov3-lite-320 | Jetson nano | float32 | 114.055ms |
mobilenet-yolov3-lite-320 | Jetson nano | float16 | 97.189 ms |
- The performance in jetson nano was not satisfactory , I will try to optimize it
see link TensorRTWrapper