Convert PyTorch model to tensorflowjs

This article is also available here.(Japanese)

At First

I want to easily run machine learning models in a browser. So I’m working on a library to run various models in tensorflowjs(see this repository). However, as you can see in the graph below, PyTorch has recently been used in 75% of the papers at major conferences, making it impossible to easily run edgy and interesting models with tensorflowjs.

State of AI Report 2020 (site)

In this situation, PINTO, an authority on model quantization, published a method for converting Pytorch to Tensorflow models at this year’s Advent Calender. Thanks for a very wonderful article.

When I tried to reproduce the model conversion using the published method, there were some stuck points in my environment, so I decided to keep this article as a working memo. So basically, I hope you will take a look at PINTO’s original article. If you get stuck, I would appreciate it if you could take a look at it just to see if it might be helpful.

[English] Converting PyTorch, ONNX, Caffe, and OpenVINO (NCHW) models to Tensorflow / TensorflowLite (NHWC) in a snap


As I mentioned above, there were some things that did not work because of my particular environment. I tried a lot of trial and error, but it didn’t work. So I finally succeeded in reproducing it using docker. In the following, I will proceed on the assumption that I will be working with docker.

Build Environment

At first, run the docker container. I used the ubuntu image which have Cuda version 10.1

$ docker run --gpus all  -v /home/whoami/docker_share:/work --name converter -ti nvidia/cuda:10.1-cudnn7-devel-ubuntu18.04

Then install related packages. I’ve added emacs and mlocate as my preference, but if you don’t need them, please omit them.

$ apt install -y python3.8 python3-pip emacs mlocate pciutils cpio git sudo curl
$ python3 -m pip install pip --upgrade
$ pip3 install tensorflow==2.3.1 --upgrade
$ pip3 install openvino2tensorflow --upgrade
$ pip3 install torch==1.7.0+cu101 torchvision==0.8.1+cu101 torchaudio==0.7.0 -f
$ pip3 install onnxruntime onnx-simplifier openvino-python --upgrade
$ pip3 install networkx defusedxml test-generator==0.1.1 tensorflow_datasets tensorflowjs

The OpenVino Toolkit does not seem to be included if you just use pip, so install it referring to the following site.

This site will ask you to download the OpenVINO toolkit package from the following site

It looks like it can be installed with pip, but when I installed it with pip, some commands were not found in my environment and I could not proceed with the work.

So I chose Web and Local Install -> Local to download the package, click on the registre & Download button and enter the required information.

Choose 2021.1 as the Version and download the Full Package. After downloading, place the file in a location that is visible from the docker container. In this case, I put it in /work/l_openvino_toolkit_p_2021.1.110.tgz.

Now, let’s install it.
When the installation starts, you will be presented with a choice. There will be various warnings, but I will skip prerequisites.

$ cp /work/l_openvino_toolkit_p_2021.1.110.tgz ./
$ tar xvfz l_openvino_toolkit_p_2021.1.110.tgz
$ cd l_openvino_toolkit_p_2021.1.110
$ ./
$ cd -

The next step is to install the dependent modules.

$ cd /opt/intel/openvino_2021/install_dependencies
$ ./
$ source /opt/intel/openvino_2021/bin/
$ cd -

That’s it!

Let’s Conversion

Now, let’s see if we can convert it well.
Let’s try to convert the Semantic Segmentation model of U²-Net by copying PINTO’s article.

Clone the U²-Net repository and download the model.

$ git clone
$ cd U-2-Net/
$ mkdir ./saved_models/u2netp/
$ curl -sc /tmp/cookie "" > /dev/null
$ CODE="$(awk '/_warning_/ {print $NF}' /tmp/cookie)"
$ curl -Lb /tmp/cookie "${CODE}&id=1rbSTGKAE-MTxBYHd-51l2hMOQPT_7EPy" -o saved_models/u2netp/u2netp.pth

Start the conversion process below.

$ export PYTHONPATH=/U-2-Net
$ SIZE=512
$ python3 /opt/intel/openvino_2021/deployment_tools/tools/model_downloader/ \
--import-module model.u2net \
--model-name U2NETP \
--input-shape 1,3,${SIZE},${SIZE} \
--weights saved_models/u2netp/u2netp.pth \
--output-file u2netp_${SIZE}x${SIZE}.onnx --input-names "x" \
--output-names "a/F.sigmoid(d0)"
ONNX check passed successfully.
$ python3 -m onnxsim u2netp_${SIZE}x${SIZE}.onnx u2netp_${SIZE}x${SIZE}_opt.onnx
Checking 2/3...
$ python3 /opt/intel/openvino_2021/deployment_tools/model_optimizer/ \
--input_model u2netp_${SIZE}x${SIZE}_opt.onnx \
--input_shape [1,3,${SIZE},${SIZE}] \
--output_dir openvino/${SIZE}x${SIZE}/FP32 \
--data_type FP32
[ SUCCESS ] Generated IR version 10 model.
[ SUCCESS ] XML file: /U-2-Net/openvino/512x512/FP32/u2netp_512x512.xml
[ SUCCESS ] BIN file: /U-2-Net/openvino/512x512/FP32/u2netp_512x512.bin
[ SUCCESS ] Total execution time: 35.76 seconds.
[ SUCCESS ] Memory consumed: 809 MB.
$ openvino2tensorflow \
--model_path openvino/${SIZE}x${SIZE}/FP32/u2netp_${SIZE}x${SIZE}_opt.xml \
--model_output_path saved_model_${SIZE}x${SIZE} \
--output_saved_model True
All the conversion process is finished! =============================================
$ tensorflowjs_converter \
--input_format=tf_saved_model \
--output_format=tfjs_graph_model \
--signature_name=serving_default \
--saved_model_tags=serve \
saved_model_${SIZE}x${SIZE} \

Now we can have the Tensorflowjs model!

Use It

Let’s use this model to actually segment the image. The result is shown below. We can see that the segmentation is working. It is a little blurry in some areas, but I think it can be improved by setting the threshold value well.
I’ve heard a lot of people say that it’s quite lightweight, but in my environment (CPU), it took 2–3 minutes. Also, with the GPU, even with a GTX1660 with 6GB of RAM, I couldn’t run it because it overflowed the memory. (I also have a 2080Ti on the same machine, but Chrome only activates the GTX1660. I don’t know how to switch it…)

In the next article, I will also try U²-Net’s Portrait Drawing. It will look like this.

I am very thirsty!!


I used the images from the following pages.



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store