Visual City Recognition
This notebook is written by dr.Seyran Khademi (s.khademi@tudelft.nl) for CIPA2019 Workshop on Computer Vision for Cultural Heritage and it is based on the paper titled: "Deep Visual City Recognition Visualization" https://arxiv.org/abs/1905.01932.
The Demo will guide you through the code to classify an image city using pretrained convolutional neural network (CNN) models. Moreover, you can inspect the visual clues that led the computer to its decision.
You can run Demo.ipynb directly using Colab platform, thus there is no need to install Jupyter notebook and other dependencies. You only need to download the pretrained CNN networks (simple.pth and simpler.pth) for classification of the city images (contact the author to access the pretrained models) and place them in your mounted google drive. You also need to place your favourite city image for classification, e.g., tokyo_test_web.jpg, in the same directory.