Abstract
This paper presents a comprehensive dataset of pelargonium species that can be used to train classification and recognition models. The provided dataset consists of 1600 images divided by 4 Pelargonium species. The collection process involved manual photography and conducting inquiries with individuals possessing imagery and knowledge regarding Pelargonium species. In addition, the paper provides a baseline method for pelargonium classification that uses Resnet50 deep convolutional neural network architecture achieving an F1-score of 73,2%. This dataset and the results of the baseline model could be used as a benchmark for further research and enhancements in the field. It is particularly useful for researchers in data science and artificial intelligence to create competitive machine learning models. The objective of the study is to create a dataset to contribute to the development of flower monitoring and flower recognition in general which can prove beneficial in the fields of botanical research, agriculture, and commerce.