/MIMO_TensorRT

Using TensorRT inference MIMO-UnetPlus on the jetson TX2, but I don't know to process output tensor.

Primary LanguageC++

MIMO_TensorRT

Using TensorRT inference MIMO-UnetPlus on the jetson TX2, but I don't know to process output tensor.

step1:

​ we need have MIMO-UnetPlus's model in onnx format. According to official guidance, we can get MIMO-UnetPlus's model in onnx format form model of the pytorch framework. GetOnnx.py has detail codes.

​ Execute the following statement in the code directory:

python3 GetOnnx.py

Note: You must have MIMO-UnetPlus's model of the pytorch framework in this directory.

step2:

​ Using trtexec generate TensorRT engine.

​ cd MIMO_Tensorrt/weights

​ trtexec --onnx='your_model'.onnx --verbose --explicitBatch --shapes=input_name:1x3x720x1280 --saveEngine='engineName'.engine

Note: Make sure your environment variables are set.

step3:

​ Begin to complie and run.

​ cd ..

​ mkdir build

​ cd build

​ cmake ..

​ make

​ ./MIMO_Tensorrt