The project idea was build on my Old Love for the Great Art and my New Love for Machine Learning. Flask App with an Image Recognition Model.
- BeautifulSoup used for web scraping, getting data and saving it on MongoDB Atlas (cloud database service)
- The Training and Testing image datasets were created for 7 artists (Nicholas Roerich, Salvador Dali, Gustav Klimt, Vincent van Gogh, Leonardo da Vinci, Pablo Picasso, Jackson Pollock). Searching and Downloading Google Images to the local disk - Thank you to https://github.com/hardikvasa/google-images-download).
ImageAI provides the most simple and powerful approach to training custom image prediction models. Thank you to https://github.com/OlafenwaMoses/ImageAI. The final Model was trained for 48 hours.
- Choose any img from imgTest folder
- Submit it and click the button 'Who is the artist?'
- Wait for predictions (less than 20 s)
- Based on prediction - collecting data about the artist from dataset and displaying it on the app.
- Embedded Tableau Story - By visually creating logical relationships between tables.
- Style Transfer in TensorFlow - use specific algorithms to manipulate a photo, style transfer attempts to identify the style of a source image on its own, with various groups of neurons working together to identify specific feature sets. Thank you to - https://github.com/lengstrom/fast-style-transfer.
- Creating GIFs with the Image module. Style Transfer had 40 steps/images to transfer Art Style to the Denver image. Those images were used for the GIFs.
- Heroku deployment - https://art-and-ml.herokuapp.com/
