2. Maestría
Permanent URI for this communityhttp://98.81.228.127/handle/20.500.12404/2
Tesis de la Escuela de Posgrado
Browse
8 results
Search Results
Item Aprendizaje profundo para transcripción de textos históricos manuscritos en español(Pontificia Universidad Católica del Perú, 2024-07-16) Choque Dextre, Gustavo Jorge; Beltrán Castañón, Cesar ArmandoEl reconocimiento de textos historicos es considerado un problema desafiante debido a los muchos factores que ´ alteran el estado de los manuscritos y la complejidad de los diferentes estilos de escritura involucrados en este tipo de documentos; en los anos recientes se han creado muchos modelos de Reconocimiento de textos manuscritos ˜ enfocados en diversos idiomas como el ingles, chino, ´ arabe y japon ´ es entre otros, sin embargo no se han ´ encontrado muchas iniciativas de reconocimiento de texto orientadas al idioma espanol debido fundamentalmente ˜ a un escasez de datasets publicos disponibles para ayudar a solucionar la problem ´ atica en dicho idioma. ´ En esta publicacion se presenta la aplicaci ´ on de t ´ ecnicas de Deep Learning basadas en una arquitectura de ´ red neuronal encoder-decoder y convoluciones compuerta Gated-CNN las cuales en los ultimos ha demostrado ´ resultados sobresalientes para resolver dicha problematica, as ´ ´ı mismo se propone la aplicacion de mecanismos de ´ Transferencia de Aprendizaje para el reconocimiento de textos historicos en espa ´ nol. Los experimentos demuestran ˜ que la aplicacion de estos m ´ etodos puede brindar resultados sobresalientes, adem ´ as la aplicaci ´ on de otras t ´ ecnicas ´ tales como Aumentacion de Datos y Modelos de Lenguaje conllevan a mejoras significativas en los resultados finales. ´ Se propone ademas el uso de un nuevo dataset de textos hist ´ oricos en espa ´ nol conformado por 1000 elementos ˜ tomados de textos historicos peruanos referentes al siglo XVIII.Item Reconocimiento de elementos de seguridad de billetes utilizando Transfer Learning(Pontificia Universidad Católica del Perú, 2021-08-12) Vera Muñoz, David; Sipiran Mendoza, Iván AnselmoLa falsificación de moneda es un problema en el país y se evidencia en informes periodísticos de incautaciones de billetes y monedas falsificadas que aparecen cada cierto tiempo a nivel nacional; por lo tanto, la necesidad de un sistema de reconocimiento de billetes y monedas es imperativo dado que a la par del crecimiento tecnológico que apoye esta tarea, también la maquinaria y tecnología utilizada para la falsificación de billetes y monedas es más accesible y costeable. La identificación de billetes y monedas falsificadas ha estado enfocada en gran medida en el procesamiento de imágenes. En el presente artículo se utiliza un modelo basado en aprendizaje por transferencia que viene teniendo buenos resultados en problemas específicos de clasificación de imágenes en la actualidad. Se ha construido un conjunto de datos con imágenes de billetes genuinos y falsificados para el entrenamiento y pruebas del modelo. Los resultados obtenidos son muy alentadores y demandan un entrenamiento más robusto con una mayor cantidad de imágenes. Asimismo con algunas mejoras en la arquitectura se podría adaptar un modelo a una aplicación móvil de manera que pueda apoyar al ciudadano de a pie en la identificación de billetes falsificados en tiempo real.Item Application on semantic segmentation with few labels in the detection of water bodies from PERUSAT-1 satellite's images(Pontificia Universidad Católica del Perú, 2020-07-02) Gonzalez Villarreal, Jessenia Margareth Marina; Beltrán Castañón, César ArmandoRemote sensing is widely used to monitor earth surfaces with the main objective of extracting information from it. Such is the case of water surface, which is one of the most affected extensions when flood events occur, and its monitoring helps in the analysis of detecting such affected areas, considering that adequately defining water surfaces is one of the biggest problems that Peruvian authorities are concerned with. In this regard, semi automatic mapping methods improve this monitoring, but this process remains a time-consuming task and into the subjectivity of the experts. In this work, we present a new approach for segmenting water surfaces from satellite images based on the application of convolutional neural networks. First, we explore the application of a U-Net model and then a transfer knowledge-based model. Our results show that both approaches are comparable when trained using an 680-labelled satellite image dataset; however, as the number of training samples is reduced, the performance of the transfer knowledge-based model, which combines high and very high image resolution characteristics, is improvedItem Object detection in videos using principal component pursuit and convolutional neural networks(Pontificia Universidad Católica del Perú, 2018-05-03) Tejada Gamero, Enrique David; Rodríguez Valderrama, Paul AntonioObject recognition in videos is one of the main challenges in computer vision. Several methods have been proposed to achieve this task, such as background subtraction, temporal differencing, optical flow, particle filtering among others. Since the introduction of Convolutonal Neural Networks (CNN) for object detection in the Imagenet Large Scale Visual Recognition Competition (ILSVRC), its use for image detection and classification has increased, becoming the state-of-the-art for such task, being Faster R-CNN the preferred model in the latest ILSVRC challenges. Moreover, the Faster R-CNN model, with minimum modifications, has been succesfully used to detect and classify objects (either static or dynamic) in video sequences; in such setup, the frames of the video are input “as is” i.e. without any pre-processing. In this thesis work we propose to use Robust PCA (RPCA, a.k.a. Principal Component Pursuit, PCP), as a video background modeling pre-processing step, before using the Faster R-CNN model, in order to improve the overall performance of detection and classification of, specifically, the moving objects. We hypothesize that such pre-processing step, which segments the moving objects from the background, would reduce the amount of regions to be analyzed in a given frame and thus (i) improve the classification time and (ii) reduce the error in classification for the dynamic objects present in the video. In particular, we use a fully incremental RPCA / PCP algorithm that is suitable for real-time or on-line processing. Furthermore, we present extensive computational results that were carried out in three different platforms: A high-end server with a Tesla K40m GPU, a desktop with a Tesla K10m GPU and the embedded system Jetson TK1. Our classification results attain competitive or superior performance in terms of Fmeasure, achieving an improvement ranging from 3.7% to 97.2%, with a mean improvement of 22% when the sparse image was used to detect and classify the object with the neural network, while at the same time, reducing the classification time in all architectures by a factor raging between 2% and 25%.Item Estudio de la posibilidad de utilizar una cámara CCD chaeleon para obtener imágenes de la fluorescencia de la vegetación a nivel de campo(Pontificia Universidad Católica del Perú, 2017-03-10) Loayza Loza, Hildo Mac Lean; Moya Marco, Ismael; Guerra Torres, Jorge AndrésLa agricultura es una actividad económica esencial para el desarrollo humano y es el sustento de millones de personas fomentando la seguridad alimentaria e impulsando la economía de países en desarrollo. Sin embargo la producción agrícola es afectada por diferentes factores abióticos (cambio climático) y bióticos (plagas y enfermedades) que disminuyen drásticamente su eficiencia. La fotosíntesis es el único mecanismo de entrada de energía de la biosfera y agentes estresantes como, por ejemplo, la escasez de agua limitan la conductancia estomática en las hojas disminuyendo la entrada de CO2 en las plantas lo que conlleva a una reducción de la fotosíntesis (Flexas et al., 2002a; Flexas y Medrano 2002a). Por estos motivos, resulta importante detectar el estrés antes que los síntomas visuales sean evidentes.Item Artificial tactile sensors for surface texture detection - analytical and numerical investigations(Pontificia Universidad Católica del Perú, 2017-02-07) Scharff, Moritz; Alencastre Miranda, Jorge Hernán; Behn, CarstenNatural vibrissae fulfill a lot of functions. Next to object distance detection and object shape recognition, the surface texture can be determined. Inspired by the natural process of surface texture detection, the goal is to adapt it by technical concepts. Modeling the vibrissa as an EulerBernoulli bending beam and the vibrissa-surface contact with respect to Coulomb's Law of Friction, the first approach is formed by the group of Steigenberger and Behn. Due to the surface contact, the vibrissa gets deformed. Initiating a linear movement of the beam support in the way that the bearn tip gets pushed, first the beam tip is sticking to the surface. The acting friction force prevents a movement of the beam tip until the static friction coeflicient is reached. The displacement of the support corresponds to changes in the acting forces and moment. Out of these changes the coeflicient of static friction can be determined. Advancing the present model, the effects of an elastic support, a conical shape of the considered beam, a natural pre-curved (stress free) beam and an inclined contact plane on the resulting forces and moments are analyzed in an analytical way, and then discussed by numerical simulations in performing parameter studies. All these special features of the beam as a tactile sensor are successfully studied. The results for the conical beam shape are only of theoretical relevance. In a next step, a quasi-static model is compared to experimental data to verify the concept. The displacement is represented by a linear, stepwise change of the support of the sensor. By image processing the deformations of the beam for every support position are analyzed. This information is compared to the simulation. The concept in principal is confirmed by the experiments.Item Desarrollo y comparación de diversos mapas de probabilidades en 3D del cáncer de próstata a partir de imágenes de histología(Pontificia Universidad Católica del Perú, 2013-12-04) Díaz Rojas, Kristians Edgardo; Castañeda Aphan, BenjamínUnderstanding the spatial distribution of prostate cancer and how it changes according to prostate specific antigen (PSA) values, Gleason score, and other clinical parameters may help comprehend the disease and increase the overall success rate of biopsies. This work aims to build 3D spatial distributions of prostate cancer and examine the extent and location of cancer as a function of independent clinical parameters. The border of the gland and cancerous regions from whole-mount histopathological images are used to reconstruct 3D models showing the localization of tumor. This process utilizes color segmentation and interpolation based on mathematical morphological distance. 58 glands are deformed into one prostate atlas using a combination of rigid, a ne, and b-spline deformable registration techniques. Spatial distribution is developed by counting the number of occurrences in a given position in 3D space from each registered prostate cancer. Finally a di erence between proportions is used to compare di erent spatial distributions. Results show that prostate cancer has a significant di erence (SD) in the right zone of the prostate between populations with PSA greater and less than 5 ng=ml. Age does not have any impact in the spatial distribution of the disease. Positive and negative capsule-penetrated cases show a SD in the right posterior zone. There is SD in almost all the glands between cases with tumors larger and smaller than 10% of the whole prostate. A larger database is needed to improve the statistical validity of the test. Finally, information from whole-mount histopathological images could provide better insight into prostate cancer.Item Automatic regularization parameter selection for the total variation mixed noise image restoration framework(Pontificia Universidad Católica del Perú, 2013-03-27) Rojas Gómez, Renán Alfredo; Rodríguez Valderrama, Paúl AntonioImage restoration consists in recovering a high quality image estimate based only on observations. This is considered an ill-posed inverse problem, which implies non-unique unstable solutions. Regularization methods allow the introduction of constraints in such problems and assure a stable and unique solution. One of these methods is Total Variation, which has been broadly applied in signal processing tasks such as image denoising, image deconvolution, and image inpainting for multiple noise scenarios. Total Variation features a regularization parameter which defines the solution regularization impact, a crucial step towards its high quality level. Therefore, an optimal selection of the regularization parameter is required. Furthermore, while the classic Total Variation applies its constraint to the entire image, there are multiple scenarios in which this approach is not the most adequate. Defining different regularization levels to different image elements benefits such cases. In this work, an optimal regularization parameter selection framework for Total Variation image restoration is proposed. It covers two noise scenarios: Impulse noise and Impulse over Gaussian Additive noise. A broad study of the state of the art, which covers noise estimation algorithms, risk estimation methods, and Total Variation numerical solutions, is included. In order to approach the optimal parameter estimation problem, several adaptations are proposed in order to create a local-fashioned regularization which requires no a-priori information about the noise level. Quality and performance results, which include the work covered in two recently published articles, show the effectivity of the proposed regularization parameter selection and a great improvement over the global regularization framework, which attains a high quality reconstruction comparable with the state of the art algorithms.