This repository demonstrates a fire detection neural network model that identifies and localizes fire in images. It uses YOLOv5 (PyTorch) for object detection, achieving 0.657 mAP@.5 on a custom dataset of 1155 annotated images. The model can provide bounding box coordinates for fire regions, enabling automated responses like sprinkler activation or drone deployment. Alternative approaches using ResNet50 for classification are also explored.