Season Classifier

A FastAI-based model to classify landscape photos into one of the four seasons of the year.

Features

• Classify images into: spring, summer, autumn, winter
• Display predicted label and per-class probabilities
• Fully reproducible pipeline

Installation / Requirements

You can install the required packages in two ways:

• Using requirements.txt:

  pip install -r requirements.txt

• Directly from the notebook:

  First cell of the notebook automatically installs the required libraries

Usage

After cloning the repository, open the notebook. It is recommended to run it from top to bottom.

Note: The notebook will generate the dataset and train the model, which may take some time.

Next, upload an image and provide its path to predict_season(learn, img_path), where:

• learn: the trained FastAI Learner object (use the learn variable from the notebook)
• img_path: path to the image file you want to classify

The model will return the prediction in the following format:

Predicted season: spring
Probabilities:
- autumn: 0.0004
- spring: 0.9992
- summer: 0.0005
- winter: 0.0000

Credits

This project is based on the fast.ai Practical Deep Learning for Coders course's lesson notebook by Jeremy Howard: Is it a Bird? Creating a model from your own data.

Images were downloaded using DuckDuckGo search.