Forest is a complex ecosystem consisting mainly of trees that buffer the earth and support a myriad of life forms. It offers a wide range of advantages, such as regulating the climate, maintaining biodiversity, providing economic opportunities, and supporting the livelihoods of millions of individuals around the world. However natural calamities like forest fires can have a devastating impact on both the forest themselves and the broader environment. Therefore, it is necessary to explore automatic detection of forest fires in order to reduce natural calamities. Researchers can better plan preventive measures and extinguishing techniques with the aid of early fire detections. This study examines the use of ensemble learning approaches for fire/smoke detection from visuals. One of the main benefits of using deep learning for early fire detection and smoke detection is its ability to recognize key features for images and to identify patterns in data. These models are trained on a large dataset of labeled fire and smoke images, allowing them to identify patterns and features that are unique to fire and smoke. A dataset consisting of 13733 images was used for training and validation. This dataset included 1102 images of fire and 12631 images of smoke, which were obtained from both videos and the internet. Of these images, 12360 were used for training and 1373 were used for validation