Person:
Forcén Carvalho, Juan Ignacio

Profile Picture

Email Address

Birth Date

Research Projects

Organizational Units

Job Title

Last Name

Forcén Carvalho

First Name

Juan Ignacio

person.page.departamento

Estadística, Informática y Matemáticas

ORCID

person.page.upna

812303

Name

Filters

2019 - 201912020 - 20213
Reset filters

Settings

Search Results

Now showing 1 - 4 of 4
  • Thumbnail Image
    PublicationOpen Access
    Nuevos métodos para la combinación de características en procesamiento de imágenes
    (2021) Forcén Carvalho, Juan Ignacio; Pagola Barrio, Miguel; Barrenechea Tartas, Edurne; Estadística, Informática y Matemáticas; Estatistika, Informatika eta Matematika
    En esta memoria estudiamos diferentes problemas abiertos entorno a la agregación de información y presentamos el estudio de nuevas técnicas de agregación de características de imágenes para mejorar el rendimiento en los problemas de clasificación y recuperación de imágenes, realizando propuestas para algoritmos de visión por computador tradicional y para modelos de Deep Learning. Concretamente, hemos propuesto un método para agregar información en problemas de clasificación multi-clase a través de ensembles ponderados. También presentamos una nueva función de pooling capaz de auto aprenderse para redes neuronales convolucionales. Centrándonos en el problema de recuperación de imágenes proponemos un esquema de agregación de características donde damos mayor importancia a los objetos relevantes de la imagen y finalmente introducimos una nueva representación de Co-Ocurrencias auto aprendible para redes neuronales convolucionales capaz de capturar la correlación espacial entre diferentes características. La memoria está dividida en dos partes: Parte I. Dedicada al planteamiento del problema, la discusión de los métodos propuestos, los resultados y las conclusiones obtenidas. Parte II. Contiene las publicaciones asociadas al estudio realizado.
  • Thumbnail Image
    PublicationOpen Access
    Learning ordered pooling weights in image classification
    (Elsevier, 2020) Forcén Carvalho, Juan Ignacio; Pagola Barrio, Miguel; Barrenechea Tartas, Edurne; Bustince Sola, Humberto; Estadística, Informática y Matemáticas; Estatistika, Informatika eta Matematika; Universidad Pública de Navarra / Nafarroako Unibertsitate Publikoa
    Spatial pooling is an important step in computer vision systems like Convolutional Neural Networks or the Bag-of-Words method. The spatial pooling purpose is to combine neighbouring descriptors to obtain a single descriptor for a given region (local or global). The resultant combined vector must be as discriminant as possible, in other words, must contain relevant information, while removing irrelevant and confusing details. Maximum and average are the most common aggregation functions used in the pooling step. To improve the aggregation of relevant information without degrading their discriminative power for image classification, we introduce a simple but effective scheme based on Ordered Weighted Average (OWA) aggregation operators. We present a method to learn the weights of the OWA aggregation operator in a Bag-of-Words framework and in Convolutional Neural Networks, and provide an extensive evaluation showing that OWA based pooling outperforms classical aggregation operators.
  • Thumbnail Image
    PublicationOpen Access
    Co-occurrence of deep convolutional features for image search
    (Elsevier, 2020) Forcén Carvalho, Juan Ignacio; Pagola Barrio, Miguel; Barrenechea Tartas, Edurne; Bustince Sola, Humberto; Estatistika, Informatika eta Matematika; Institute of Smart Cities - ISC; Estadística, Informática y Matemáticas
    Image search can be tackled using deep features from pre-trained Convolutional Neural Networks (CNN). The feature map from the last convolutional layer of a CNN encodes descriptive information from which a discriminative global descriptor can be obtained. We propose a new representation of co-occurrences from deep convolutional features to extract additional relevant information from this last convolutional layer. Combining this co-occurrence map with the feature map, we achieve an improved image representation. We present two different methods to get the co-occurrence representation, the first one based on direct aggregation of activations, and the second one, based on a trainable co-occurrence representation. The image descriptors derived from our methodology improve the performance in very well-known image retrieval datasets as we prove in the experiments.
  • Thumbnail Image
    PublicationOpen Access
    Moderate deviation and restricted equivalence functions for measuring similarity between data
    (Elsevier, 2019) Altalhi, A. H.; Forcén Carvalho, Juan Ignacio; Pagola Barrio, Miguel; Barrenechea Tartas, Edurne; Bustince Sola, Humberto; Takáč, Zdenko; Estatistika, Informatika eta Matematika; Institute of Smart Cities - ISC; Estadística, Informática y Matemáticas
    In this work we study the relation between moderate deviation functions, restricted dissimilarity functions and restricted equivalence functions. We use moderate deviation functions in order to measure the similarity or dissimilarity between a given set of data. We show an application of moderate deviate functions used in the same way as penalty functions to make a final decision from a score matrix in a classification problem.