Deep Generative AI Based on Denoising Diffusion Probabilistic Models for Applications in Image Processing
Descripción del Articulo
Denoising diffusion probabilistic models (DDPMs) have demonstrated significant potential in addressing complex image processing challenges. This paper explores the application of DDPMs in three different areas: reconstruction of remote sensing imagery affected by cloud cover, reconstruction of facia...
Autores: | , , , , |
---|---|
Formato: | artículo |
Fecha de Publicación: | 2024 |
Institución: | Universidad de Lima |
Repositorio: | Revistas - Universidad de Lima |
Lenguaje: | inglés español |
OAI Identifier: | oai:revistas.ulima.edu.pe:article/7389 |
Enlace del recurso: | https://revistas.ulima.edu.pe/index.php/Interfases/article/view/7389 |
Nivel de acceso: | acceso abierto |
Materia: | machine learning reconstruction segmentation face and gesture recognition remote sensing aprendizaje automático reconstrucción segmentación reconocimiento facial y gestual teledetección |
id |
REVULIMA_47cce91ef2eb24181807fc9b3b404e8c |
---|---|
oai_identifier_str |
oai:revistas.ulima.edu.pe:article/7389 |
network_acronym_str |
REVULIMA |
network_name_str |
Revistas - Universidad de Lima |
repository_id_str |
|
dc.title.none.fl_str_mv |
Deep Generative AI Based on Denoising Diffusion Probabilistic Models for Applications in Image Processing IA generativa profunda basada en modelos de difusión de desenfoque probabilístico para aplicaciones en procesamiento de imágenes |
title |
Deep Generative AI Based on Denoising Diffusion Probabilistic Models for Applications in Image Processing |
spellingShingle |
Deep Generative AI Based on Denoising Diffusion Probabilistic Models for Applications in Image Processing Bezerra, Emili Silva machine learning reconstruction segmentation face and gesture recognition remote sensing aprendizaje automático reconstrucción segmentación reconocimiento facial y gestual teledetección |
title_short |
Deep Generative AI Based on Denoising Diffusion Probabilistic Models for Applications in Image Processing |
title_full |
Deep Generative AI Based on Denoising Diffusion Probabilistic Models for Applications in Image Processing |
title_fullStr |
Deep Generative AI Based on Denoising Diffusion Probabilistic Models for Applications in Image Processing |
title_full_unstemmed |
Deep Generative AI Based on Denoising Diffusion Probabilistic Models for Applications in Image Processing |
title_sort |
Deep Generative AI Based on Denoising Diffusion Probabilistic Models for Applications in Image Processing |
dc.creator.none.fl_str_mv |
Bezerra, Emili Silva Leher, Quefren Oliveira Alves, Uendel Diego da Silva Paixão, Thuanne Alvarez, Ana Beatriz |
author |
Bezerra, Emili Silva |
author_facet |
Bezerra, Emili Silva Leher, Quefren Oliveira Alves, Uendel Diego da Silva Paixão, Thuanne Alvarez, Ana Beatriz |
author_role |
author |
author2 |
Leher, Quefren Oliveira Alves, Uendel Diego da Silva Paixão, Thuanne Alvarez, Ana Beatriz |
author2_role |
author author author author |
dc.subject.none.fl_str_mv |
machine learning reconstruction segmentation face and gesture recognition remote sensing aprendizaje automático reconstrucción segmentación reconocimiento facial y gestual teledetección |
topic |
machine learning reconstruction segmentation face and gesture recognition remote sensing aprendizaje automático reconstrucción segmentación reconocimiento facial y gestual teledetección |
description |
Denoising diffusion probabilistic models (DDPMs) have demonstrated significant potential in addressing complex image processing challenges. This paper explores the application of DDPMs in three different areas: reconstruction of remote sensing imagery affected by cloud cover, reconstruction of facial images with occluded areas, and segmentation of bodies of water from remote sensing imagery. Inpainting involves filling in missing regions in images, while DDPMs act as data generators capable of synthesizing information that alings coherently with the context of the original data. Inspired by the inpainting technique, the RePaint approach was adapted and applied to reconstruction tasks. The WaterSegDiff approach, which uses a diffusion model as a backbone, was employed for the segmentation task. To illustrate the model’s behavior and provide examples of the tasks, experiments were carried out with both qualitative and quantitative evaluations. The qualitative results show the model’s ability to generate data for reconstruction and segmentation. Quantitatively, metrics such as MSE, PSNR, SSIM, IoU, PA and F1 score highlight the model’s proficient performance in image processing tasks. In this scenario, DDPMs have proved to be a promising tool for high-quality data reconstruction, enabling the hallucination of image regions with high visual coherence and facilitating applications in various areas, such as environmental monitoring, facial recognition, water resource mapping, among others. |
publishDate |
2024 |
dc.date.none.fl_str_mv |
2024-12-26 |
dc.type.none.fl_str_mv |
info:eu-repo/semantics/article info:eu-repo/semantics/publishedVersion |
format |
article |
status_str |
publishedVersion |
dc.identifier.none.fl_str_mv |
https://revistas.ulima.edu.pe/index.php/Interfases/article/view/7389 10.26439/interfases2024.n020.7389 |
url |
https://revistas.ulima.edu.pe/index.php/Interfases/article/view/7389 |
identifier_str_mv |
10.26439/interfases2024.n020.7389 |
dc.language.none.fl_str_mv |
eng spa |
language |
eng spa |
dc.relation.none.fl_str_mv |
https://revistas.ulima.edu.pe/index.php/Interfases/article/view/7389/7463 https://revistas.ulima.edu.pe/index.php/Interfases/article/view/7389/7464 |
dc.rights.none.fl_str_mv |
https://creativecommons.org/licenses/by/4.0 info:eu-repo/semantics/openAccess |
rights_invalid_str_mv |
https://creativecommons.org/licenses/by/4.0 |
eu_rights_str_mv |
openAccess |
dc.format.none.fl_str_mv |
application/pdf text/html |
dc.publisher.none.fl_str_mv |
Universidad de Lima |
publisher.none.fl_str_mv |
Universidad de Lima |
dc.source.none.fl_str_mv |
Interfases; No. 020 (2024); 71-93 Interfases; Núm. 020 (2024); 71-93 Interfases; n. 020 (2024); 71-93 1993-4912 10.26439/interfases2024.n020 reponame:Revistas - Universidad de Lima instname:Universidad de Lima instacron:ULIMA |
instname_str |
Universidad de Lima |
instacron_str |
ULIMA |
institution |
ULIMA |
reponame_str |
Revistas - Universidad de Lima |
collection |
Revistas - Universidad de Lima |
repository.name.fl_str_mv |
|
repository.mail.fl_str_mv |
|
_version_ |
1841719317103640576 |
spelling |
Deep Generative AI Based on Denoising Diffusion Probabilistic Models for Applications in Image ProcessingIA generativa profunda basada en modelos de difusión de desenfoque probabilístico para aplicaciones en procesamiento de imágenesBezerra, Emili SilvaLeher, Quefren OliveiraAlves, Uendel Diego da SilvaPaixão, ThuanneAlvarez, Ana Beatrizmachine learningreconstructionsegmentationface and gesture recognitionremote sensingaprendizaje automáticoreconstrucciónsegmentaciónreconocimiento facial y gestualteledetecciónDenoising diffusion probabilistic models (DDPMs) have demonstrated significant potential in addressing complex image processing challenges. This paper explores the application of DDPMs in three different areas: reconstruction of remote sensing imagery affected by cloud cover, reconstruction of facial images with occluded areas, and segmentation of bodies of water from remote sensing imagery. Inpainting involves filling in missing regions in images, while DDPMs act as data generators capable of synthesizing information that alings coherently with the context of the original data. Inspired by the inpainting technique, the RePaint approach was adapted and applied to reconstruction tasks. The WaterSegDiff approach, which uses a diffusion model as a backbone, was employed for the segmentation task. To illustrate the model’s behavior and provide examples of the tasks, experiments were carried out with both qualitative and quantitative evaluations. The qualitative results show the model’s ability to generate data for reconstruction and segmentation. Quantitatively, metrics such as MSE, PSNR, SSIM, IoU, PA and F1 score highlight the model’s proficient performance in image processing tasks. In this scenario, DDPMs have proved to be a promising tool for high-quality data reconstruction, enabling the hallucination of image regions with high visual coherence and facilitating applications in various areas, such as environmental monitoring, facial recognition, water resource mapping, among others.Los denoising diffusion probabilistic models (DDPMs) han mostrado un potencial significativo en la resolución de problemas complejos de procesamiento de imágenes. Este estudio explora el uso de DDPMs en tres aplicaciones diferentes, incluyendo la reconstrucción de imágenes de teledetección en zonas con nubosidad, la reconstrucción de imágenes faciales con regiones ocluidas y la segmentación de masas de agua a partir de imágenes de teledetección. El inpainting consiste en rellenar las regiones omitidas en las imágenes, mientras que los DDPM actúan como generadores de datos capaces de sintetizar información coherente con el contexto de los datos originales. En este contexto, tomando la técnica de inpainting como inspiración, se adaptó el enfoque RePaint y se aplicó a tareas de reconstrucción. Para la tarea de segmentación se utilizó la técnica WaterSegDiff, que también utiliza un modelo de difusión como backbonner. Para ilustrar el comportamiento del modelo y ejemplificar las tareas, se realizaron experimentos cuya performance se evaluó cualitativa y cuantitativamente. Los resultados de las evaluaciones cualitativas muestran la capacidad del modelo para generar datos para la reconstrucción y la segmentación. Cuantitativamente, las métricas MSE, PSNR, SSIM, IoU, PA y F1-Score indican un hábil desempeño de los modelos en tareas de procesamiento de imágenes. En este escenario, los DDPMs han demostrado ser una herramienta prometedora para la reconstrucción de datos de alta calidad, permitiendo la alucinación de regiones de imágenes con alta coherencia visual y aplicaciones en diversas áreas, tales como monitoreo ambiental, reconocimiento facial, mapeo de recursos hídricos, entre otros. Os Denoising Diffusion Probabilistic Models (DDPMs) têm demonstrado um potencial significativo na resolução de problemas complexos de processamento de imagem. Neste estudo foi explorada a utilização dos DDPMs em três aplicações distintas, incluindo, reconstrução de imagens de sensoriamento remoto em áreas com cobertura de nuvens, reconstrução de imagens faciais com regiões ocluídas e segmentação de corpos d'água a partir de imagens de sensoriamento remoto. A técnica Inpainting consiste em preencher regiões faltantes em imagens, por outro lado os DDPMs atuam como geradores de dados capazes de sintetizar informações coerentes com o contexto do dado original. Nesse contexto, inspirados na técnica inpainting, a abordagem RePaint foi adaptada e aplicada para as tarefas de reconstrução. Já para a tarefa de segmentação foi utilizada a técnica WaterSegDiff que também utiliza um modelo de difusão como backbonner. Para ilustrar o comportamento do modelo e exemplificar as tarefas foram realizados experimentos cujo desempenho foi avaliado qualitativa e quantitativamente. Os resultados das avaliações qualitativas evidenciam a capacidade do modelo em gerar dados para reconstrução e segmentação. Quantitativamente as métricas MSE, PSNR, SSIM, IoU, PA e F1-Score indicam o ótimo desempenho dos modelos em tarefas de processamento de imagens. Nesse cenário, os DDPMs demonstraram ser uma ferramenta promissora para a reconstrução de dados com alta qualidade, permitindo alucinação de imagens com alta coerência visual e aplicações em diversas áreas, como monitoramento ambiental, reconhecimento facial, mapeamento de recursos hídricos, entre outros.Universidad de Lima2024-12-26info:eu-repo/semantics/articleinfo:eu-repo/semantics/publishedVersionapplication/pdftext/htmlhttps://revistas.ulima.edu.pe/index.php/Interfases/article/view/738910.26439/interfases2024.n020.7389Interfases; No. 020 (2024); 71-93Interfases; Núm. 020 (2024); 71-93Interfases; n. 020 (2024); 71-931993-491210.26439/interfases2024.n020reponame:Revistas - Universidad de Limainstname:Universidad de Limainstacron:ULIMAengspahttps://revistas.ulima.edu.pe/index.php/Interfases/article/view/7389/7463https://revistas.ulima.edu.pe/index.php/Interfases/article/view/7389/7464https://creativecommons.org/licenses/by/4.0info:eu-repo/semantics/openAccessoai:revistas.ulima.edu.pe:article/73892025-05-02T12:58:55Z |
score |
12.924651 |
Nota importante:
La información contenida en este registro es de entera responsabilidad de la institución que gestiona el repositorio institucional donde esta contenido este documento o set de datos. El CONCYTEC no se hace responsable por los contenidos (publicaciones y/o datos) accesibles a través del Repositorio Nacional Digital de Ciencia, Tecnología e Innovación de Acceso Abierto (ALICIA).
La información contenida en este registro es de entera responsabilidad de la institución que gestiona el repositorio institucional donde esta contenido este documento o set de datos. El CONCYTEC no se hace responsable por los contenidos (publicaciones y/o datos) accesibles a través del Repositorio Nacional Digital de Ciencia, Tecnología e Innovación de Acceso Abierto (ALICIA).