Deep Learning Architecture

Image classifier that will predict what style of architecture a building is influenced by. It is my intent to use this as a benchmark to try different modesls and libraries for training.

Problem Statement

The goal of this project is to create an image classifier that will accurately predict what style of architecture a building is influenced by. This would be very useful to any professionals in the real estate market who are looking to either label or search for property programmatically.

Data Set:

Architectural Style Classification using MLLR. It contains measurements from 25 architecture styles for 5000 buildings. URL: https://sites.google.com/site/zhexuutssjtu/projects/arch

File size : 782 MB Format of data file: images This dataset contains 25 folders which represent the 25 classes of architecture styles: ['Achaemenid architecture', 'American craftsman style', 'American Foursquare architecture', 'Ancient Egyptian architecture', 'Art Deco architecture', 'Art Nouveau architecture', 'Baroque architecture', 'Bauhaus architecture', 'Beaux-Arts architecture', 'Byzantine architecture', 'Chicago school architecture', 'Colonial architecture', 'Deconstructivism', 'Edwardian architecture', 'Georgian architecture', 'Gothic architecture', 'Greek Revival architecture', 'International style', 'Novelty architecture', 'Palladian architecture', 'Postmodern architecture', 'Queen Anne architecture', 'Romanesque architecture', 'Russian Revival architecture', 'Tudor Revival architecture']

Description of Hardware

I developed and executed this application on a custom linux box that i made running ubuntu on an Intel® CoreTM i5-6200 CPU processor. The laptop is equipped with 32 gigabytes of RAM and a dedicated graphics processing unit.

Description of Software

This project is developed using keras, tensorflow and Python (version 3.6.3). I have been using Python on the Anaconda platform. Anaconda is a very useful and convenient Python Data Science Platform. It installs a variety of Python Packages of which we will be using Pandas Dataframes, Numpy, Scipy, Matplotlib and Jupyter Notebook. One can download the Anaconda platform from the following URL: https://www.anaconda.com/download/