算法部署-使用TensorRT部署RefineNet图像分割算法-优质算法部署项目实战.zip

# RefineNet_TensorRT ## environment Ubuntu1604 Pytorch 0.41 TensorRT5.1.5 OpenCV3.4.8 ## Build 1. configure your TensorRT path in CMakeLists.txt 2. make: ``` mkdir build && cd build cmake .. make -j8 ``` 3. ``` ./RefineNet s float16(float32) refinenet.engine ../vid/demo.mp4 ../refinenet.onnx ``` 4. serialize the engine from onnx model: ``` ./RefineNet s float16(float32) refinenet.engine ../vid/face.mp4 ../refinenet.onnx ``` 5. deserialize the engine and infer: ``` ./RefineNet infer float16 refinenet.engine ../vid/face.mp4 ``` <img src="./image/d94be52120f2aa2cfbd7c12f10817b04.jpeg" style="zoom:25%;" /> <img src="./image/Screenshot from 2020-08-16 13-39-14.png" style="zoom: 50%;" /> ## Performance | Model | FPS | | ------- | ---- | | Pytorch | 5 | | FP32 | 27 | | FP16 | 33 |