A Deep Learning approach to detect forest fires from images or videos. The model is a Sequential model, made up of a 8 layered Convolutional Neural Network with five Convolutional and Max Pooling layers. The model has been trained on a total of 5000 images, where 2500 images belong to Fire class and 2500 belong to Non Fire class. The overall accuracy of the model, on the validation data, is 93.90%
- To test out the model, open the ipynb file to explore the cells.
- The train/test images are available on my custom dataset on Kaggle: https://www.kaggle.com/mohnishsaiprasad/forest-fire-images
- Create a subfolder named Video_Test_Data and procure some Forest Fire videos to test.
- Open the script.py file under Code folder and run it on the termial.
- If a Fire has been detected in the video, the program sends an email to the address specified in the script.