This project aims to detect and classify forest fires using deep learning techniques, specifically the VGG-19 convolutional neural network model. The model is trained to analyze images and accurately predict whether they contain signs of a forest fire or not.
- Transfer Learning: Utilizes pre-trained weights from the VGG-19 model on ImageNet for feature extraction.
- Data Augmentation: Enhances the training dataset through image augmentation techniques for improved model generalization.
- Binary Classification: Classifies images into two categories: fire or non-fire, facilitating quick decision-making.
- Hyperparameter Tuning: Flexibility to experiment with hyperparameters such as epochs, batch size, and learning rate for optimal performance.
- Validation Metrics: Monitors and displays validation loss and accuracy during training for model evaluation.
- Loss Curves: Visualizes the training and validation loss curves over epochs to understand the model's learning progress.
- Predictions and Evaluation: Generates predictions on test and validation sets, providing insights into model performance.
The dataset used for training and evaluation has been sourced from Kaggle. You can find the original dataset here.