2025-03-072025-03-072021-04-09SOUSA, Rafael Nascimento de. Aplicação de rede neural convolucional na classificação de azulejos históricos da cidade de Belém do Pará. Orientador: Carlos Gustavo Resque dos Santos. 2021. 42 f. Trabalho de Conclusão de Curso (Bacharelado em Ciência da Computação) – Faculdade de Computação, Instituto de Ciências Exatas e Naturais, Universidade Federal do Pará, Belém, 2021. Disponível em:. Acesso em:.https://bdm.ufpa.br/jspui/handle/prefix/7783The cultural heritage of a city is of great importance for the maintenance and enhancement of its history. Innovative technologies such as Augmented Reality and Computer Vision can be used to emphasize these legacies that the city itself carries in an attractive and playful way. In the city of Belém do Pará, tiles are an important example of the cultural heritage present in the city, which goes back to the moments and characteristics of its foundation. The recognition of images, for example, can facilitate the search for historical information about a certain tile in a faster way, since with just one photo of the tile, your information would be returned to the user, thus facilitating the search and access to information on the tiles, in addition to their identification. In view of what has been exposed, this Course Conclusion Paper presents a prototype that uses Convolutional Neural Networks (CNN) to classify, through images, the historical tiles of Belém. CNN’s training used three image bases, being the the first composed of images that contained tiles and other elements of the environment (walls, doors, building facades, etc.), the second composed of images with a focus on tile patterns and the third composed of photos taken from the computer screen, from of second base images, and once created the bases are combined into a final base used in training the model. Twelve different types of tiles were used, representing the classes to be recognized by CNN. Thus, after training, the tflite model (Tensorflow Lite) with the classes of tiles is generated, to be used in the image classification task on a mobile device. With the model and the classes defined, an application was created for this conclusion work in which the user takes a picture of a tile and the four classes are returned with the best accuracy and the historical information of the classified tiles. For the image classification task, a module was created that gathers the necessary functions for the image classification task and that can be reused in other projects.Acesso AbertoReconhecimento de ImagemHerança culturalAzulejoRedes neuraisConvolucionaisTensorflow liteImage recognitionCultural heritageConvolutional neural networksTensorflow liteCNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAOTrabalho de Curso - Graduação - Monografia