Neural Style Transfer is a task of transferring style of one image to another. It is accomplished by using features of some pretrained model. In this project, the employed model was used for localized stylization of specific shape of the product image.GrabCut Algorithm and Region-Confined Stylization are being employed for this research project.

• System Requirements
• Installation
• Run the Application
• Usage
• License
• Reference

Operating system: Windows 10
Programming Langauge: Python 3
External Python Libraries: Please refer to the requirements.txt

pip3 install -r requirements.txt

To run the application, please first download the VGG-19 model into the main project directory and simply click the run.bat to start.

The following screenshot shows the completed execution result of the GUI of the Python application after clickling the run.bat

The sample of original images and style images are stored in the content folder and style folder respectively, the final stylized image is stored in the images folder as result.jpg

This project is licensed under MIT License.

• An Interactive Neural Network-Based System for Confined Stylization of Product Design

Man-Kit, Tang.”An Interactive Neural Network-Based System for Confined Stylization of Product Design”. In: Proceedings of 12th International Conference on Design and Semantics of Form and Movement (DesForM 2023), p.102-108, July 2023.

• Constrained Neural Style Transfer for Decorated Logo Generation
• Grab Cut Algorithm Implementation