In this article, our main goal is to implement efficient fire detection using the Keras library and discuss Artificial Intelligence, Deep Learning, and the Application of Artificial Intelligence in Unmanned Aerial Vehicles. In this work, our code is written in Python using different libraries like Keras, NumPy, OpenCV, Matplotlib, etc. The training will involve two classes: fire images and neutral images. The total number of training images is 1004, with the same number of validation images in the fire and neutral directories, totaling 294 images. This dataset is custom-made and a total of 1298 images were used in the dataset. CNN architecture was used in this fire detection project. Accuracy, loss, validation loss, and validation accuracy values obtained in this article are also shown in graphs. The results we achieved in this project are satisfactory for a custom dataset. The project we executed has been tested on video, cameras, and images, yielding positive outcomes.