The diverse plant and animal species that make up forests are essential components of ecosystems that have developed over time to co-exist. Wildfires, which may originate naturally, inadvertently by humans, or as a result of lightning strikes, frequently pose a threat to such ecosystems. Early wildfire detection is essential for preserving people, property, and resources. The approaches for localizing and categorizing images of forests under three varied domains of Fires, No Fire and Smoke using various data augmentation and pre-processing techniques are reviewed in this study. Additionally, this work makes use of employing the dataset for detecting wildfires in forests, or just the presence of smoke which resolves the classification issue of distinguishing between photographs with and without a fire. This work proposes an effectiveness study of various pre-trained deep learning (DL) neural networks namely Inceptionv3, VGG19, VGG16 and ResNet50. Inceptionv3 outshone all the other models, with regards to performance over datasets without CLAHE 95.12% with CLAHE 97.63% accuracy.