The development of technology in the field of weather information is needed especially for all aspects of life. To recognize, study, and detect weather conditions that occur, classification techniques with the help of artificial intelligence are needed. The classification model used is a convolutional neural network (CNN) with a modified LeNet-5 architecture. The purpose of this study is to test the performance of the model for the classification of sunny, cloudy, cloudy and rainy weather conditions, as well as to determine the resulting accuracy and its application. With this model. The image size used is 224x224, batch size 32, learning rate 0.0001 and trained with 50 epochs. In the model training process, 8 different scenarios were created involving augmentation and no augmentation techniques, as well as the use of one of the callbacks functions in the form of earlystopping. The CNN model that uses augmentation and earlystopping with a patience value of 5 produces the best performance because it achieves an accuracy of up to 94%. The model is implemented on a locally hosted website and produces predictions that match the weather conditions that occur