Building Type Classifier
Through my attempts to learn about Machine Learning and Deep Learning, this is a simple classifier to classify images of residential and industrial areas accordingly. This was, in fact, a challenge from the Udacity Bertelsmann Deep Learning scholarship. The Kaggle competition link can be found here.
Both models and Notebooks used the PyTorch framework.
Models
In this repo, there are two notebooks: one pertraining to transfer learning and another with an CNN (Convolutional Neural Network). The accuracy of both, as it's shown in the following table, are almost identical.
Model | Overall Accuracy | Residential Accuracy | Industrial Accuracy |
---|---|---|---|
ResNet | 92% | 97% | 87% |
Own CNN | 91% | 90% | 92% |