keras2TensorRTConverter

keras2TensorRTConverter is a Python program that allow you to convert keras .hdf5 models to optimized tensorflow .pb files.

Tested on Ubuntu 18.04 Tensorflow 1.14

Download

git clone https://github.com/Acuno41/keras2TensorRTConverter.git

Installation

Use the pip to install requirements.

pip3 install requirements.txt

Usage

Download Keras pre-trained resNet50 imagenet model with:

python3 weights/downloadModel.py

To convert .hdf5 file to optimized .pb file run:

python3 keras2TensorRTConverter.py --modelDir weights/ --modelName resNet50_imagenet.hdf5

to make inference with the optimized tensorrt model run: (The important point should be paid attention to the input and output layer names in the previous script output.)

python3 predictTensorrt.py --modelPath weights/resNet50_imagenet.pb --imagePath images/container.jpeg --inputLayerName input_1 --outputLayerName fc1000/Softmax

Contributing

Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.

Please make sure to update tests as appropriate.