ONNX Image Style Transfer Model (Flask App + Docker)

Introduction

In this project, we use pretrained onnx image style transfer models to build a flask app packaged in docker for implememting image conversions. The goal is to demonstrate the ease of running models packaged as onnx forms compared to models in their native forms. Models packaged in onnx forms runs fatser and are easier to deploy than models used in their native forms.

This app is using Flask 1.1.2 and Python 3.8.8.

Files

model.py

contains the python onnx model for transforming image

• python3 model.py to run

app.py

contains the flask app of the model

• python3 app.py to run

templates

contains html templats for running the flask app

• index.html

Dockerfile

built docker image from ubuntu18.04. To build cocker image and run it use the commands below

• docker build .
• docker run -p

requirements.txt

dependencies for running the app

• to generate the requirements file use the command below
• pipreqs /"Working Directory"/

Tech stack

• Ubuntu 18.04
• Python 3.8.8
• Flask 1.1.2
• Docker

All computations were done on ubuntu 18.04

Running this app

You'll need to have Docker installed. It's available on Windows, macOS and most distros of Linux.

Check it out in a browser:

Visit http://localhost:8000 in your favorite browser.