Our jetson environment follows jkjung's jetson-nano
To install the TensorFlow on Nvidia Jetson devices, please refer to here. Protobuf 3.8.0 also has to be installed as follows:
$ mkdir ${HOME}/project
$ cd ${HOME}/project
$ git clone https://github.com/jkjung-avt/jetson_nano.git
$ cd jetson_nano
$ ./install_protobuf-3.8.0.sh
$ source ${HOME}/.bashrc
To install dependencies and build the TensorRT Engine, please enter the commands as follow.
$ cd ${HOME}/project
$ git clone https://github.com/jkjung-avt/tensorrt_demos.git
$ cd ${HOME}/project/tensorrt_demos/ssd
$ ./install_pycuda.sh
$ sudo pip3 install onnx==1.4.1
$ cd ${HOME}/project/tensorrt_demos/plugins
$ make
To use this runtime module, you need to clone this repository in your local storage and install dependencies that are described in ‘requirements.txt’.
$ git clone https://github.com/Nota-NetsPresso/NetsPresso-Model-Searcher-Runtime.git
$ pip3 install -r requirements.txt
To execute the ‘runtime.py’, you can use an example model with sample images that we provide, which can be accessed on this link below.
Example command:
$ python3 runtime.py --model ./runtime_example/example.tflite --image_folder ./runtime_example/ --classes ./runtime_example/data.yaml --mode save