Video Demo:

Code for the Model

Code for the Web App Intergration

Trained Model Ready For Use

Model Performance on test dataset

Model Prediction Accuracy: 90%

Confusion Matrix:

Model Architecture

Used Keras to Implemented the VGG16 Architecture, with 2 nodes at the output layer to tailor to the binary classification task

Model training results:

Model Limitations:

The training data set only contains images of trees/forests not on fire for the "Non_fire" training set. Inputting any images that is not a tree that is not on fire, like cars and planes, the model may falsely predict "fire". Could train on a more diverse dataset to obtain better predictions

Data Set

• Cleaned up, balanced dataset used in the model training

• Original Data Set from Kaggle