The goal of this mini-project is to carry-out a simple end-to-end image classification project. The task is to train a CNN model for classifying an image of an art painting as one of the two genres - portrait or landscape. The input dataset consists of 20 000 images of each class obtained by scraping the WikiArt art library (code for this can be found here).
The notebook is structured to mirror a typical modeling workflow:
- Download and sort input images from Google Drive
- Split input images into train, validation, and test sets
- Define image preprocessing and TF datasets
- Train a baseline CNN model
- Visually assess the model
- Tune a pretrained model
The whole project is coded in portland.ipynb.