Every year, approximately 70,000 wildfires occur around the world. Forest fires can have a variety of negative effects on forest cover, soil, tree development, vegetation, and general flora and fauna. Fires destroy several hectares of forest and leave behind ash, rendering it unfit for vegetation growth. Animal habitats are destroyed by the heat created by the fire. Preventing fire disasters and saving people's lives and property requires early detection of fires before they become catastrophic. These forest fires must be detected and put out as early as possible. In this project a computer vision based deep learning model is proposed that can detect and alert in-case of forest fires. The model which is used here is called MobileNet Architecture which employs depth-wise separable convolutions and has greater advantages than using other currently existing models. The model is trained from a dataset that was scrapped from the internet. The computer vision model is built utilizing the packages in Python 3 such as NumPy, OpenCV, Pillow, Matplotlib, and TensorFlow. This method is created by training and testing the model to identify failures and establish recovery strategies. As a result, these computer vision-based systems can provide a warning at an early stage of fire, which is critical for forest fire early warning.