Classification of Architectural Styles Using Neural Networks

Understanding features that define a building's architectural style

Code from a neural network that was trained on 90,000 images of buildings to classify a building's architectural style, and two dimension reduction techniques that take the output of the neural network prior to the classification layer and tries to find faithful representations of those high dimensional points to lower dimensional space (variational autoencoder and t-distributed stochastic neighbor embedding)

---

Data used are Google images of buildings labelled as 9 different architectural styles: art deco, art nouveau, Chinese, gothic, modernist, neoclassicm, renaissance, romanesque, Russian

---

Tools Used:

Python
Keras
Pandas
Matplotlib
scikit-learn
d3.js
Flask

---

Check out my blog Classification of Architectural Styles Using Neural Networks