{"id":{"repo_id":"lima","oai_identifier":"oai:repositorio.ulima.edu.pe:20.500.12724/24823"},"canonical_url":"https://search.dev.ndltd.org/etd/lima/oai:repositorio.ulima.edu.pe:20.500.12724/24823","repository":{"repo_id":"lima","name":"Universidad de Lima","base_url":"https://repositorio.ulima.edu.pe/oai/request"},"display":{"title":"A robust gan-based model for low-light image enhancement and human detection on mobile devices","abstract":"Detecting people in low-light environments makes it difficult to recognize them with the naked eye, and this can be affected by weather conditions or the time of day. This reduces the reliability of people detection models, as they attempt to detect a shape that is altered by darkness. When the lighting decreases, the image changes, in addition to the noise generated by the lack of clarity. To address this problem, a Generative Adversarial Network (GAN) model is proposed to reconstruct a dark image by eliminating the details that prevented visualization. This is achieved through a concept called ratio-log, which integrates lighting inputs. The results obtained were 25.15 PSNR and 0.909 SSIM using the LOLv1 dataset, and a different result of 28.04 PSNR and 0.905 SSIM using the surveillance-oriented dataset, both showing good results. Furthermore, testing using the YOLOv12 model revealed certain errors when attempting to identify people in dark environments with noise. Integrating image enhancements showed an improvement from 0.44 to 0.78, revealing individuals previously hidden in dark images. Additionally, the GAN model was integrated into an Android application built with a clean architecture, enabling image enhancement within the mobile device. Finally, a survey of IT professionals yielded moderately positive results, confirming its accessibility for users and for future developers who wish to modify it in new research.","abstract_html":"Detecting people in low-light environments makes it difficult to recognize them with the naked eye, and this can be affected by weather conditions or the time of day. This reduces the reliability of people detection models, as they attempt to detect a shape that is altered by darkness. When the lighting decreases, the image changes, in addition to the noise generated by the lack of clarity. To address this problem, a Generative Adversarial Network (GAN) model is proposed to reconstruct a dark image by eliminating the details that prevented visualization. This is achieved through a concept called ratio-log, which integrates lighting inputs. The results obtained were 25.15 PSNR and 0.909 SSIM using the LOLv1 dataset, and a different result of 28.04 PSNR and 0.905 SSIM using the surveillance-oriented dataset, both showing good results. Furthermore, testing using the YOLOv12 model revealed certain errors when attempting to identify people in dark environments with noise. Integrating image enhancements showed an improvement from 0.44 to 0.78, revealing individuals previously hidden in dark images. Additionally, the GAN model was integrated into an Android application built with a clean architecture, enabling image enhancement within the mobile device. Finally, a survey of IT professionals yielded moderately positive results, confirming its accessibility for users and for future developers who wish to modify it in new research.","abstract_has_math":false,"creators":["Camayo Ramos, Sebastian Enrique"],"institution":"Universidad de Lima","degree_name":"Ingeniero de Sistemas","degree_level":"Titulo profesional","degree_discipline":"Ingeniería de Sistemas","degree_department":null,"school":null,"contributors":[],"advisors":["Escobedo Cárdenas, Edwin Jonathan"],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026","date_published":"2026","updated_at":"2026-07-24T02:50:08Z","subjects":[],"languages":["eng"],"rights":[],"rights_urls":["https://purl.org/coar/access_right/c_abf2","https://creativecommons.org/licenses/by-nc-sa/4.0/"],"identifier_entries":[]},"links":{"outbound_url":"https://hdl.handle.net/20.500.12724/24823","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Escobedo Cárdenas, Edwin Jonathan"]},{"key":"dc:creator","label":"Author","values":["Camayo Ramos, Sebastian Enrique"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2026-06-11T14:19:45Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2026-06-11T14:19:45Z"]},{"key":"dc:date.issued","label":"Date","values":["2026"]},{"key":"dc:publisher","label":"Institution","values":["Universidad de Lima"]},{"key":"dc:type","label":"Dc Type","values":["https://purl.org/coar/resource_type/c_7a1f"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Ingeniería de Sistemas"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Titulo profesional"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Ingeniero de Sistemas"]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["Universidad de Lima. Facultad de Ingeniería"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["eng"]},{"key":"dc:rights","label":"Dc Rights","values":["https://purl.org/coar/access_right/c_abf2"]},{"key":"dc:rights.uri","label":"Rights URI","values":["https://creativecommons.org/licenses/by-nc-sa/4.0/"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/20.500.12724/24823"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Detecting people in low-light environments makes it difficult to recognize them with the naked eye, and this can be affected by weather conditions or the time of day. This reduces the reliability of people detection models, as they attempt to detect a shape that is altered by darkness. When the lighting decreases, the image changes, in addition to the noise generated by the lack of clarity. To address this problem, a Generative Adversarial Network (GAN) model is proposed to reconstruct a dark image by eliminating the details that prevented visualization. This is achieved through a concept called ratio-log, which integrates lighting inputs. The results obtained were 25.15 PSNR and 0.909 SSIM using the LOLv1 dataset, and a different result of 28.04 PSNR and 0.905 SSIM using the surveillance-oriented dataset, both showing good results. Furthermore, testing using the YOLOv12 model revealed certain errors when attempting to identify people in dark environments with noise. Integrating image enhancements showed an improvement from 0.44 to 0.78, revealing individuals previously hidden in dark images. Additionally, the GAN model was integrated into an Android application built with a clean architecture, enabling image enhancement within the mobile device. Finally, a survey of IT professionals yielded moderately positive results, confirming its accessibility for users and for future developers who wish to modify it in new research.","La detección de personas en un entorno de baja iluminación no permite su reconocimiento a simple vista y esto puede verse afectado por las condiciones climáticas o el momento del día. Esto reduce la confiabilidad en los modelos de detección de personas, ya que intentan detectar una forma que se ve modificada por la oscuridad. Cuando la iluminación se reduce se hace un cambio en la imagen añadiendo ruido que se genera al no tener una imagen clara. Ante esta problemática se propone un modelo de Redes Generativas Antagónicas (GAN) que permita una reconstrucción de la imagen eliminando así la oscuridad que impiden ver los detalles en ella junto con un concepto llamado ratio-log que integra los inputs de iluminación para que la mejora sea más suave. Los mejores resultados que se obtuvieron fueron de 25.15 PSNR y 0.909 SSIM usando el dataset LOLv1 y con otro distinto de 28.04 PSNR y 0.905 SSIM usando el dataset de surveillance-oriented siendo buenos resultados en ambos. Además, mediante un testeo usando el modelo YOLOv12 se pudo reconocer ciertos errores al intentar identificar a las personas en ambientes oscuros junto con ruido, porque al integrar las mejoras en la imagen se pudo notar que hubo una mejora al revelar a personas que no se notaban en las imágenes con oscuridad. También, el modelo GAN fue integrado dentro de una aplicación Android construida con una arquitectura limpia habilitando la mejora de imágenes dentro del dispositivo móvil. Finalmente, se hizo una encuesta con profesionales en el campo de TI dejando un resultado medianamente positivo confirmando así que su accesibilidad es adecuada para las personas y para los próximos desarrolladores que quieran modificarlo en nuevas investigaciones."]},{"key":"dc:format","label":"Dc Format","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["A robust gan-based model for low-light image enhancement and human detection on mobile devices"]}]}],"canonical_facts":{"dc:contributor.advisor":["Escobedo Cárdenas, Edwin Jonathan"],"dc:creator":["Camayo Ramos, Sebastian Enrique"],"dc:date.accessioned":["2026-06-11T14:19:45Z"],"dc:date.available":["2026-06-11T14:19:45Z"],"dc:date.issued":["2026"],"dc:description.abstract":["Detecting people in low-light environments makes it difficult to recognize them with the naked eye, and this can be affected by weather conditions or the time of day. This reduces the reliability of people detection models, as they attempt to detect a shape that is altered by darkness. When the lighting decreases, the image changes, in addition to the noise generated by the lack of clarity. To address this problem, a Generative Adversarial Network (GAN) model is proposed to reconstruct a dark image by eliminating the details that prevented visualization. This is achieved through a concept called ratio-log, which integrates lighting inputs. The results obtained were 25.15 PSNR and 0.909 SSIM using the LOLv1 dataset, and a different result of 28.04 PSNR and 0.905 SSIM using the surveillance-oriented dataset, both showing good results. Furthermore, testing using the YOLOv12 model revealed certain errors when attempting to identify people in dark environments with noise. Integrating image enhancements showed an improvement from 0.44 to 0.78, revealing individuals previously hidden in dark images. Additionally, the GAN model was integrated into an Android application built with a clean architecture, enabling image enhancement within the mobile device. Finally, a survey of IT professionals yielded moderately positive results, confirming its accessibility for users and for future developers who wish to modify it in new research.","La detección de personas en un entorno de baja iluminación no permite su reconocimiento a simple vista y esto puede verse afectado por las condiciones climáticas o el momento del día. Esto reduce la confiabilidad en los modelos de detección de personas, ya que intentan detectar una forma que se ve modificada por la oscuridad. Cuando la iluminación se reduce se hace un cambio en la imagen añadiendo ruido que se genera al no tener una imagen clara. Ante esta problemática se propone un modelo de Redes Generativas Antagónicas (GAN) que permita una reconstrucción de la imagen eliminando así la oscuridad que impiden ver los detalles en ella junto con un concepto llamado ratio-log que integra los inputs de iluminación para que la mejora sea más suave. Los mejores resultados que se obtuvieron fueron de 25.15 PSNR y 0.909 SSIM usando el dataset LOLv1 y con otro distinto de 28.04 PSNR y 0.905 SSIM usando el dataset de surveillance-oriented siendo buenos resultados en ambos. Además, mediante un testeo usando el modelo YOLOv12 se pudo reconocer ciertos errores al intentar identificar a las personas en ambientes oscuros junto con ruido, porque al integrar las mejoras en la imagen se pudo notar que hubo una mejora al revelar a personas que no se notaban en las imágenes con oscuridad. También, el modelo GAN fue integrado dentro de una aplicación Android construida con una arquitectura limpia habilitando la mejora de imágenes dentro del dispositivo móvil. Finalmente, se hizo una encuesta con profesionales en el campo de TI dejando un resultado medianamente positivo confirmando así que su accesibilidad es adecuada para las personas y para los próximos desarrolladores que quieran modificarlo en nuevas investigaciones."],"dc:format":["application/pdf"],"dc:identifier.uri":["https://hdl.handle.net/20.500.12724/24823"],"dc:language.iso":["eng"],"dc:publisher":["Universidad de Lima"],"dc:rights":["https://purl.org/coar/access_right/c_abf2"],"dc:rights.uri":["https://creativecommons.org/licenses/by-nc-sa/4.0/"],"dc:title":["A robust gan-based model for low-light image enhancement and human detection on mobile devices"],"dc:type":["https://purl.org/coar/resource_type/c_7a1f"],"thesis:degree_discipline":["Ingeniería de Sistemas"],"thesis:degree_level":["Titulo profesional"],"thesis:degree_name":["Ingeniero de Sistemas"],"thesis:institution_name":["Universidad de Lima. Facultad de Ingeniería"]},"updated_at":"2026-07-24T02:50:08Z"}