In this work, we presented an image classification of dragonflies. The dragonflies are one of the oldest winged species (approximately 300 million years old) that have not changed significantly over time. Species composition changes can indicate global changes in environment under human interference. The samples used for training classifier were gathered by photographers using different cameras and therefore are not homogenous. In total, we took and labeled roughly 1000 images of dragonflies belonging into 179 classes distinguished by species and sex. To expand our dataset for the needs of deep learning methods, we have manually added 20 000 of images from internet image search.The dataset was used to train a deep convolutional neural network (SqueezeNet) that was pre-trained on ImageNet database. The model was finetuned on the created dataset. In the study, we had to deal with issues of the over-fitting, data pre-processing and augmentation, different data distributions, unbalanced data, and others. The classifier will serve as a baseline for mobile application for online dragonflies species identification. The application will classify dragonflies species on a fine scale, to provide an alternative to existing solutions (iNaturalist dataset, etc.). Gathered data from application users will be used for further research.