🎨 Neural Style Transfer
A practical deep learning project that transforms ordinary photos into stylised artwork. It uses a VGG19-based neural style transfer pipeline to combine the content of a photograph with the visual style of famous paintings, producing artistic images with distinctive textures, colours, and patterns.
Designed for creative AI experiments and image processing workflows, the system can run through either a simple web interface or a command-line pipeline. Users can upload their own images, choose from a range of artistic styles, adjust the strength of the effect, and generate high-quality outputs locally on their own machine.
✨ Key Features
- Transforms photos into artwork using neural style transfer
- Supports famous painting styles and custom style images
- Offers adjustable style strength and processing settings
- Includes both a web interface and command-line workflow
- Runs locally for private image processing
- Produces high-quality stylised output images
🔍 How It Works
The system uses a pre-trained VGG19 network to separate image content from artistic style. It preserves the structure of the original photo while optimising the output image to match the textures, colours, and visual patterns of the selected artwork, producing a blended result over multiple iterations.
🎯 Practical Uses
This project is useful for creative coding, digital art experiments, AI image processing demonstrations, and exploring how deep learning can be applied beyond standard classification tasks. It provides a clear and practical example of how neural networks can be used to generate visually compelling artistic transformations from everyday photographs.
Tech stack: Python, PyTorch, VGG19, OpenCV, Pillow, and Flask.