Forest Fire Image Classification

In this project, we trained a Convolutional Neural Network on Images with and without Forest Fires. Our final F1 Score on the test data set was 0,964. An additional visual test on 20 images proofed the model to be confident in its classification decisions. We suggest the model to be further tested in the field.

Description

In the 01_EDA Notebook we perform a light exploratory analysis on the dataset.
In the 02_Modelling notebook we perform Deep Learning Image Classification on the Kaggle Forest Fire Images Data Set. We train a Convolutional NN and optimize its F1 Score. We also utilize Data Augmentation, Dropout and the Keras-Tuner Framework to optimize our Model.

Getting Started

Dependencies

• Numpy
• Matplotlib
• Tensorflow
• Keras

Authors

Contributors names and contact info

Julia Albrecht
@juliaalbrecht

Acknowledgments

Inspiration, code snippets, etc.

• https://www.tensorflow.org/tutorials/images/classification
• https://keras.io/examples/vision/image_classification_from_scratch/