Art.ai

NeuralStyleTransfer

Description

This project is a Deep Learning based Web application, that helps demonstrate and simulate the concept of Neural Style Transfer. Neural style transfer is an optimization technique used to take two images — a content image and a style reference image (such as an artwork by a famous painter) — and blend them together so the output image looks like the content image, but “painted” in the style of the style reference image. The project is based on A Neural Algorithm of Artistic Style paper using TensorFlow & Keras.

Getting Started

1. Clone this repo.
2. Make sure all necessary dependencies are installed.
3. Dive into the project folder & run python run.py in your terminal. The flask server will start at port 5000, by default.
4. Upload content & style images.
5. Configure the settings - choose between TensorFlow-Hub (default) or TensorFlow Manual Training.
6. Sit tight & monitor the progress.

Screenshots

Major Tech Stack

• Python
• TensorFlow
• Flask
• Socket.io
• MaterializeCSS

Future Scope

• Provide option for selecting curated style images.
• Improve speed of the manual training by using single feed-forward pass only.
• Provide option for sharing images.
• Applying NST on videos.

References

• TensorFlow - Neural Style Transfer
• PyImageSearch - Neural Style Transfer with OpenCV
• Medium Posts-
  • Chi-Feng Wang - A Basic Introduction to Separable Convolutions
  • Sahil Singla - Experiments on different loss configurations for style transfer
  • Sahil Singla - Practical techniques for getting style transfer to work
  • Kailash Ahirwar - Important resources if you are working with Neural Style Transfer or Deep Photo Style Transfer
  • Connor Shorten - Towards Fast Neural Style Transfer
• StackOverflow Blogs-
  • How can I access my localhost from my Android device?
  • Configure Flask dev server to be visible across the network