Week 1 — Image Transformation according to Art Style
Introduction
Hi folks! Today, we would like to talk about our machine learning capstone project. In this project, we aim to implement an image-style transformation model by considering the art styles of different artists.
Many artist’s works from the past to the present can be reachable as digital data. This opportunity gives us a huge advantage to work on artist’s styles. We purpose to develop an application of machine learning to artist identification for a style transformation. In this way, we will transform real images into artist-styled images. We will classify styles of paintings, and we will try to train our model with this knowledge for style transformation of real images to styled images. Let's talk about business!
About Data
We plan to use the Kaggle data which includes 50 different artists with their arts. Also, nationalities, genres, and other stuff about the arts are placed in this dataset. However, we work on the image data directly.
You can click here to reach the data.
Work Plan
We made some research about image transformation according to style. We found that the GAN framework gives the most accurate results on this problem. We plan to use the GAN as a base for our works. The framework gives the transformed images, but there is a missing, which is the evaluation part.
The log-likelihood method is used in many generative models to evaluate accuracy. In statistics, the likelihood function (often simply called the likelihood) measures the goodness of fit of a statistical model to a sample of data for given values of the unknown parameters.
Thanks for reading, see you in the next post 😎
