personal-web-image-assets/machine-learning/photo2vangogh/app
personal-web-image-assets/machine-learning/photo2vangogh/img1-origin
personal-web-image-assets/machine-learning/photo2vangogh/img1-transform
personal-web-image-assets/machine-learning/photo2vangogh/img2-origin
personal-web-image-assets/machine-learning/photo2vangogh/img2-transform
personal-web-image-assets/machine-learning/photo2vangogh/img3-origin
personal-web-image-assets/machine-learning/photo2vangogh/img3-transform
personal-web-image-assets/machine-learning/photo2vangogh/img4-origin
personal-web-image-assets/machine-learning/photo2vangogh/img4-transform
personal-web-image-assets/machine-learning/photo2vangogh/img5-origin
personal-web-image-assets/machine-learning/photo2vangogh/img5-transform
Introduction
Style transfer app allow you to transform a real photo into a vangogh panting style. The result is combination of real photo with vangogh style.

Motivation & Purpose
Learning a particular skill of panting style then apply it on drawing of beautiful scene or object would be a time consuming and difficult process. Not everyone can do it. Hand over this task to artificial intelligent can shorten this process.
This artifical intelligent learn the style of painting from vangogh and then transform a real photo into the style.

More
The model was trained with dataset from Kaggle.
The model use CycleGAN to achieve style transfer for real photo.
The model's size had been reduced by using DepthwiseConv2D instead of Conv2D.
Frameworks were used to build this app:
    🧰 Tensorflow : Deep learning framework for the model
    🧰 Streamlit : Frontend app

Note
From time to time this app is not able to complete transfer style properly due to dataset was limited and hardware restriction during model training.

Website