Indonesia is a country with a tropical climate that greatly affects agriculture. Flowering plants are estimated to account for 25% of species in Indonesia; there are 416 families, 13,164 genera, and 295,383 species of flowering plants. Classification of profit types is a time- and knowledge-intensive job. Convolutional Neural Network (CNN) has revolutionized the field of computer vision by improving the accuracy of image, text, voice, and video recognition. This research is focused on developing a CNN model for Indonesian flower images by optimizing hyperparameters combined with a grid search algorithm and default parameters, as well as comparing two different CNN architectures, namely VGG16 and MobileNetV2. This research aims to improve the classification accuracy of Indonesian flower images by optimizing hyperparameters. The results of CNN research with hyperparameters combined with a grid search algorithm and using data augmentation resulted in MobileNetV2 as the best model. Grid search is designed to get the best value of each parameter. The performance of the grid search algorithm can produce an optimal combination of parameters, with a test accuracy of 89.62%..