Desarrollo de una aplicación para la detección del tipo de mascota utilizando machine learnig

Abstract

In this thesis an application is developed for the detection of the type of pet using machine learning techniques, focusing specifically on the recognition of dogs, cats and birds. The main goal is to develop and run a system that can accurately recognize these types of animals from images. To achieve this, several key stages have been followed: data collection and preprocessing, the design of the (CNN) model architecture, the training of the model with a wide and diverse variety of data, and the evaluation of its performance using standard metrics in the field of machine learning. During the data collection stage, images have been obtained from different sources to ensure their diversity and representativeness. Preprocessing has included normalization and data augmentation techniques to improve the robustness of the model. The creation of the model structure has been based on best practices and recent studies in the field of CNNs, optimizing both the depth and complexity of the network to balance accuracy and computational efficiency. The model was trained using a hardware infrastructure suitable for intensive processing, achieving significant accuracy in the classification of pet images. A distinctive feature of this project is the implementation of the application on a Raspberry Pi 4, demonstrating the feasibility of running machine learning models on low-cost, resourceconstrained devices. System integration on the Raspberry Pi 4 includes model optimization to ensure adequate performance in terms of inference period and memory usage. The results obtained demonstrate the effectiveness of the proposed model, achieving high precision and recall rates in the identification of dogs, cats and birds. Finally, possible applications of the system in areas such as veterinary medicine, the pet trade, and consumer technology are discussed, as well as the study's findings and future directions to improve and expand the scope of the system.