Identification of the weather conditions is crucial for daily human activities, agricultural production, and safe transportation. The increased number of surveillance and personal cameras provide images and valuable information for recognizing the weather conditions in different regions at any moment. However, accurate models are required to avoid misclassification, which might cause severe situations. The advancements in deep learning techniques and transfer learning have become highly effective in using existing knowledge of the models, extracting different levels of features, and performing robust predictions on image data. This paper implements a pre-trained network, EfficientNetV2 B0, in order to classify four weather conditions, such as cloudy, shine, sunrise, and rain, from the still images. The model is trained with a 5-fold cross-validation and hold-out method in different experiments, and a comprehensive analysis is performed. The results show that the EfficientNetV2 B0 model can predict weather conditions with up to 98.66% accuracy and is superior to the models in recent studies.