Skip to content

Image Retrieval system of Traditional Chinese Painting Based on Convolutional Neural Network

Notifications You must be signed in to change notification settings

Muhongfan/IMAGE-PROCESSING-BASED-ON-DEEP-LEARNING

Repository files navigation

Image Retrieval system of Traditional Chinese Painting Based on Convolutional Neural Network

#ELG5378_IP_Project

Image Processing and Image Communication

Traditional Chinese painting can be traced back nearly five thousand years and has created a large number of stunning and unique art forms. With the development of digital technology, digital art is considered one of the most convenient and effective ways to help educate and enhance the public's aesthetic taste. However, due to the long history of Chinese painting, each artist in different periods and dynasties has established various preferences and techniques for their own works. This has created great difficulty for the public to understand the various styles of art works and the connection between artists and art history periods. The expected use case of this application is to provide a search engine for similar art pieces to visitors, students, or art curators who are interested in learning about the history and related information behind the artwork while touring an exhibition hall or studying art in a university setting, but due to certain reasons, lacking direct access to name or author information about the artwork. They can use this application to obtain information about the history and related information of art works and similar pieces, but only when they are unable to directly obtain the name or author information for the artwork.

In this project, it:

  1. developed an automatic analysis system to analyze subtle differences between different art works and output the genre and artist.
  2. considered data augmentation for oversampling, and applied the VGG-16 architecture to extract features from input images,
  3. compared the extracted features in a feature vector database.
  4. established the system based on Django fronted framework.
  5. finally, presented the two highest similarity results on the website system.

Slides can be seen on https://drive.google.com/file/d/1nNwGCQ1O3GO2_GUzbWMzRUh4N5XxSvci/view?usp=share_link

My contribution: all the coding for website setup, modelling; result analysis and future work

About

Image Retrieval system of Traditional Chinese Painting Based on Convolutional Neural Network

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published