This paper considers the development and training of a neural network model for the identification of the species and quantitative composition of pollen, which will subsequently be used to determine the botanical origin of honey and quantitative calculations of pollen grains contained in a certain mass of honey. The main purpose of the study is to create a model that can effectively distinguish the morphology of pollen grains present in honey, determine their quantitative and qualitative composition, which will improve product quality control, as well as identify its botanical and geographical origin. For this purpose, similar works on pollen classification were analyzed. Pwtorch was chosen as the framework for creating a neural network model, which provides the possibility of detailed configuration of the model. The result of the work is a trained model capable of classifying pollen grains.
Keywords: neural networks, classification task, pollen classification, convolutional neural networks, PyTorch
In today's world, facial recognition is becoming an increasingly important and relevant task. With the development of technology and the increasing amount of data, the need for reliable, accurate and efficient face recognition systems increases. Neural networks demonstrate high efficiency in solving computer vision problems and have great potential for improving existing mathematical models of face recognition. This paper is devoted to the study of methods for human face recognition, the Viola-Jones algorithm will be discussed in detail, which, which can be applied in the task of face recognition using neural networks. It will also analyse techniques for training deep learning models using libraries that also use the Viola-Jones algorithm and describe an algorithm for using the trained model in an API that can be used in desktop and mobile applications.
Keywords: biometric identification, human face recognition, mathematical models, face recognition methods, deep learning, convolutional neural networks, tensorflow