MULTI-TASKING DEEP LEARNING FOR OBJECT RECOGNITION (Q3159224): Difference between revisions
Jump to navigation
Jump to search
(Removed claim: summary (P836): DEEP LEARNING (‘DEEP LEARNING’) HAS BECOME THE TECHNIQUE PAR EXCELLENCE IN THE FIELD OF AUTOMATIC LEARNING APPLIED TO COMPUTER VISION. BASED ON ITS SUCCESS IN THE CLASSIFICATION OF IMAGES, DEEP LEARNING HAS SURPASSED (IN TERMS OF PERFORMANCE) OTHER TECHNIQUES OF THE SAME SCOPE AND IS BEING USED TODAY BY THE MAJORITY OF APPLICATIONS IN VISION BY COMPUTER, INCLUDING THE DETECTION OF ‘SALIENCY’, THE DETECTION OF OBJECTS, VISUAL TRACKING, IMAGE PR...) |
(Created claim: summary (P836): DEEP LEARNING (‘DEEP LEARNING’) HAS BECOME THE TECHNIQUE PAR EXCELLENCE IN THE FIELD OF AUTOMATIC LEARNING APPLIED TO COMPUTER VISION. BASED ON ITS SUCCESS IN THE CLASSIFICATION OF IMAGES, DEEP LEARNING HAS SURPASSED (IN TERMS OF PERFORMANCE) OTHER TECHNIQUES OF THE SAME SCOPE AND IS BEING USED TODAY BY THE MAJORITY OF APPLICATIONS IN VISION BY COMPUTER, INCLUDING THE DETECTION OF ‘SALIENCY’, THE DETECTION OF OBJECTS, VISUAL TRACKING, IMAGE PROC...) |
||||||||||||||
| Property / summary | |||||||||||||||
DEEP LEARNING (‘DEEP LEARNING’) HAS BECOME THE TECHNIQUE PAR EXCELLENCE IN THE FIELD OF AUTOMATIC LEARNING APPLIED TO COMPUTER VISION. BASED ON ITS SUCCESS IN THE CLASSIFICATION OF IMAGES, DEEP LEARNING HAS SURPASSED (IN TERMS OF PERFORMANCE) OTHER TECHNIQUES OF THE SAME SCOPE AND IS BEING USED TODAY BY THE MAJORITY OF APPLICATIONS IN VISION BY COMPUTER, INCLUDING THE DETECTION OF ‘SALIENCY’, THE DETECTION OF OBJECTS, VISUAL TRACKING, IMAGE PROCESSING AND THE AUTOMATION OF THE PROCESS OF DESCRIBING THE CONTENTS OF THE IMAGE. The EXIT achieved by the PROFUND NEURONAL networks is due to the existence of very large data bases and the development of new GPUS necessities for the process of basic data._x000D__x000D_ The multitasking apprehending (MTL — BY YOUR SIGLAS IN ENGLISH) it’s a very good technic that has been established by the SCIENTIFIC COMMUNITY OF AUTOMATIC LEARNING. THIS TECHNIQUE CONTEMPLATES SEVERAL TASKS IN A JOINT WAY WITH THE AIM OF TAKING ADVANTAGE OF THE SIMILARITIES THEY SHARE. AS A RESULT, AN IMPROVEMENT IN PERFORMANCE CAN BE ACHIEVED THAN IF EACH TASK WERE CONSIDERED SEPARATELY. YOUR APPLICATION IS SUITABLE FOR ANY PROBLEM WHERE THERE ARE A NUMBER OF RELATED TASKS TO LEARN AND/OR WHEN THERE IS A SMALL DATASET TO BE ABLE TO LEARN A PARTICULAR TASK. MTL was initiated by the capacity of the human being to improve the apprehending process of a court if it is carried out in a way that is linked to other territories that have been linked to, at the same time as the same time, an individual way._x000D_ _x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x005F the OBJECTIVE OF THIS MEMORY IS TO PROPONER AND STUDY A THEORIC FRAMEWORK AS A COMMUNITY INCLUDING THE TWO CONCEPTS: DEEP LEARNING AND MTL. THE INITIATIVE OF THIS PROJECT IS BASED ON THE EXISTENCE OF PROMISING INITIAL RESULTS IN THIS DIRECTION. OUR WORKING HYPOTHESIS IS: DEEP NEURAL NETWORKS CAN BENEFIT FROM THE MULTITASKING LEARNING STRATEGY, I.E. MULTI-TASK-TRAINED NETWORKS INCREASE THEIR PERFORMANCE COMPARED TO SINGLE-TASK-TRAINED NETWORKS. IN ADDITION, MULTITASKING LEARNING CAN BE VERY USEFUL WHEN THE DATASET FOR A TASK IS NOT LARGE ENOUGH. IN THIS CASE THE ABSENCE OF DATA IN ONE TASK IS COMPENSATED BY THE DATA OF THE OTHER REMAINING TASKS. FOR THIS REASON, WE INTEND TO STUDY IN DETAIL THE DEEP MULTITASK LEARNING (DMTL) AS PART OF THIS PROJECT. WE INTEND TO DEFINE A GENERAL FRAMEWORK FOR THE DESIGN OF DMTL ARCHITECTURES AND OPTIMAL STRATEGIES FOR THEIR TRAINING. IN ADDITION, THE EFFECT OF DECOMPENSATED DATASETS TO LEARN DMTL NETWORKS WILL BE STUDIED. SPECIFICALLY, DMTL NETWORKS WILL BE APPLIED TO SEVERAL APPLICATIONS INCLUDING OBJECT DETECTION, CLASSIFICATION OF SCENES, DETECTION OF THE ‘SALIENCY’, AUTOMATION OF THE PROCESS OF DESCRIBING THE CONTENTS OF THE IMAGE AND VISUAL MONITORING (RESEARCH THEMES IN WHICH THE LAMP GROUP IS ALREADY INVOLVED BUT WHICH WILL NOW BE CONSIDERED FROM THE POINT OF VIEW OF A SINGLE TASK). (English) | |||||||||||||||
| Property / summary: DEEP LEARNING (‘DEEP LEARNING’) HAS BECOME THE TECHNIQUE PAR EXCELLENCE IN THE FIELD OF AUTOMATIC LEARNING APPLIED TO COMPUTER VISION. BASED ON ITS SUCCESS IN THE CLASSIFICATION OF IMAGES, DEEP LEARNING HAS SURPASSED (IN TERMS OF PERFORMANCE) OTHER TECHNIQUES OF THE SAME SCOPE AND IS BEING USED TODAY BY THE MAJORITY OF APPLICATIONS IN VISION BY COMPUTER, INCLUDING THE DETECTION OF ‘SALIENCY’, THE DETECTION OF OBJECTS, VISUAL TRACKING, IMAGE PROCESSING AND THE AUTOMATION OF THE PROCESS OF DESCRIBING THE CONTENTS OF THE IMAGE. The EXIT achieved by the PROFUND NEURONAL networks is due to the existence of very large data bases and the development of new GPUS necessities for the process of basic data._x000D__x000D_ The multitasking apprehending (MTL — BY YOUR SIGLAS IN ENGLISH) it’s a very good technic that has been established by the SCIENTIFIC COMMUNITY OF AUTOMATIC LEARNING. THIS TECHNIQUE CONTEMPLATES SEVERAL TASKS IN A JOINT WAY WITH THE AIM OF TAKING ADVANTAGE OF THE SIMILARITIES THEY SHARE. AS A RESULT, AN IMPROVEMENT IN PERFORMANCE CAN BE ACHIEVED THAN IF EACH TASK WERE CONSIDERED SEPARATELY. YOUR APPLICATION IS SUITABLE FOR ANY PROBLEM WHERE THERE ARE A NUMBER OF RELATED TASKS TO LEARN AND/OR WHEN THERE IS A SMALL DATASET TO BE ABLE TO LEARN A PARTICULAR TASK. MTL was initiated by the capacity of the human being to improve the apprehending process of a court if it is carried out in a way that is linked to other territories that have been linked to, at the same time as the same time, an individual way._x000D_ _x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x005F the OBJECTIVE OF THIS MEMORY IS TO PROPONER AND STUDY A THEORIC FRAMEWORK AS A COMMUNITY INCLUDING THE TWO CONCEPTS: DEEP LEARNING AND MTL. THE INITIATIVE OF THIS PROJECT IS BASED ON THE EXISTENCE OF PROMISING INITIAL RESULTS IN THIS DIRECTION. OUR WORKING HYPOTHESIS IS: DEEP NEURAL NETWORKS CAN BENEFIT FROM THE MULTITASKING LEARNING STRATEGY, I.E. MULTI-TASK-TRAINED NETWORKS INCREASE THEIR PERFORMANCE COMPARED TO SINGLE-TASK-TRAINED NETWORKS. IN ADDITION, MULTITASKING LEARNING CAN BE VERY USEFUL WHEN THE DATASET FOR A TASK IS NOT LARGE ENOUGH. IN THIS CASE THE ABSENCE OF DATA IN ONE TASK IS COMPENSATED BY THE DATA OF THE OTHER REMAINING TASKS. FOR THIS REASON, WE INTEND TO STUDY IN DETAIL THE DEEP MULTITASK LEARNING (DMTL) AS PART OF THIS PROJECT. WE INTEND TO DEFINE A GENERAL FRAMEWORK FOR THE DESIGN OF DMTL ARCHITECTURES AND OPTIMAL STRATEGIES FOR THEIR TRAINING. IN ADDITION, THE EFFECT OF DECOMPENSATED DATASETS TO LEARN DMTL NETWORKS WILL BE STUDIED. SPECIFICALLY, DMTL NETWORKS WILL BE APPLIED TO SEVERAL APPLICATIONS INCLUDING OBJECT DETECTION, CLASSIFICATION OF SCENES, DETECTION OF THE ‘SALIENCY’, AUTOMATION OF THE PROCESS OF DESCRIBING THE CONTENTS OF THE IMAGE AND VISUAL MONITORING (RESEARCH THEMES IN WHICH THE LAMP GROUP IS ALREADY INVOLVED BUT WHICH WILL NOW BE CONSIDERED FROM THE POINT OF VIEW OF A SINGLE TASK). (English) / rank | |||||||||||||||
Normal rank | |||||||||||||||
| Property / summary: DEEP LEARNING (‘DEEP LEARNING’) HAS BECOME THE TECHNIQUE PAR EXCELLENCE IN THE FIELD OF AUTOMATIC LEARNING APPLIED TO COMPUTER VISION. BASED ON ITS SUCCESS IN THE CLASSIFICATION OF IMAGES, DEEP LEARNING HAS SURPASSED (IN TERMS OF PERFORMANCE) OTHER TECHNIQUES OF THE SAME SCOPE AND IS BEING USED TODAY BY THE MAJORITY OF APPLICATIONS IN VISION BY COMPUTER, INCLUDING THE DETECTION OF ‘SALIENCY’, THE DETECTION OF OBJECTS, VISUAL TRACKING, IMAGE PROCESSING AND THE AUTOMATION OF THE PROCESS OF DESCRIBING THE CONTENTS OF THE IMAGE. The EXIT achieved by the PROFUND NEURONAL networks is due to the existence of very large data bases and the development of new GPUS necessities for the process of basic data._x000D__x000D_ The multitasking apprehending (MTL — BY YOUR SIGLAS IN ENGLISH) it’s a very good technic that has been established by the SCIENTIFIC COMMUNITY OF AUTOMATIC LEARNING. THIS TECHNIQUE CONTEMPLATES SEVERAL TASKS IN A JOINT WAY WITH THE AIM OF TAKING ADVANTAGE OF THE SIMILARITIES THEY SHARE. AS A RESULT, AN IMPROVEMENT IN PERFORMANCE CAN BE ACHIEVED THAN IF EACH TASK WERE CONSIDERED SEPARATELY. YOUR APPLICATION IS SUITABLE FOR ANY PROBLEM WHERE THERE ARE A NUMBER OF RELATED TASKS TO LEARN AND/OR WHEN THERE IS A SMALL DATASET TO BE ABLE TO LEARN A PARTICULAR TASK. MTL was initiated by the capacity of the human being to improve the apprehending process of a court if it is carried out in a way that is linked to other territories that have been linked to, at the same time as the same time, an individual way._x000D_ _x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x005F the OBJECTIVE OF THIS MEMORY IS TO PROPONER AND STUDY A THEORIC FRAMEWORK AS A COMMUNITY INCLUDING THE TWO CONCEPTS: DEEP LEARNING AND MTL. THE INITIATIVE OF THIS PROJECT IS BASED ON THE EXISTENCE OF PROMISING INITIAL RESULTS IN THIS DIRECTION. OUR WORKING HYPOTHESIS IS: DEEP NEURAL NETWORKS CAN BENEFIT FROM THE MULTITASKING LEARNING STRATEGY, I.E. MULTI-TASK-TRAINED NETWORKS INCREASE THEIR PERFORMANCE COMPARED TO SINGLE-TASK-TRAINED NETWORKS. IN ADDITION, MULTITASKING LEARNING CAN BE VERY USEFUL WHEN THE DATASET FOR A TASK IS NOT LARGE ENOUGH. IN THIS CASE THE ABSENCE OF DATA IN ONE TASK IS COMPENSATED BY THE DATA OF THE OTHER REMAINING TASKS. FOR THIS REASON, WE INTEND TO STUDY IN DETAIL THE DEEP MULTITASK LEARNING (DMTL) AS PART OF THIS PROJECT. WE INTEND TO DEFINE A GENERAL FRAMEWORK FOR THE DESIGN OF DMTL ARCHITECTURES AND OPTIMAL STRATEGIES FOR THEIR TRAINING. IN ADDITION, THE EFFECT OF DECOMPENSATED DATASETS TO LEARN DMTL NETWORKS WILL BE STUDIED. SPECIFICALLY, DMTL NETWORKS WILL BE APPLIED TO SEVERAL APPLICATIONS INCLUDING OBJECT DETECTION, CLASSIFICATION OF SCENES, DETECTION OF THE ‘SALIENCY’, AUTOMATION OF THE PROCESS OF DESCRIBING THE CONTENTS OF THE IMAGE AND VISUAL MONITORING (RESEARCH THEMES IN WHICH THE LAMP GROUP IS ALREADY INVOLVED BUT WHICH WILL NOW BE CONSIDERED FROM THE POINT OF VIEW OF A SINGLE TASK). (English) / qualifier | |||||||||||||||
point in time: 12 October 2021
| |||||||||||||||
Revision as of 16:30, 12 October 2021
Project Q3159224 in Spain
| Language | Label | Description | Also known as |
|---|---|---|---|
| English | MULTI-TASKING DEEP LEARNING FOR OBJECT RECOGNITION |
Project Q3159224 in Spain |
Statements
19,723.0 Euro
0 references
39,446.0 Euro
0 references
50.0 percent
0 references
30 December 2016
0 references
29 December 2019
0 references
CENTRO DE VISION POR COMPUTADOR
0 references
41°29'27.71"N, 2°8'15.00"E
0 references
08266
0 references
EL APRENDIZAJE PROFUNDO ('DEEP LEARNING') SE HA CONVERTIDO EN LA TECNICA POR EXCELENCIA EN EL AMBITO DEL APRENDIZAJE AUTOMATICO APLICADO A LA VISION POR COMPUTADOR. TOMANDO COMO REFERENCIA SU EXITO EN LA CLASIFICACION DE IMAGENES, EL APRENDIZAJE PROFUNDO HA SUPERADO (EN TERMINOS DE RENDIMIENTO) A OTRAS TECNICAS DEL MISMO AMBITO Y ESTA SIENDO UTILIZADA HOY EN DIA POR LA MAYORIA DE LAS APLICACIONES EN VISION POR COMPUTADOR, INCLUYENDO LA DETECCION DE 'SALIENCY', LA DETECCION DE OBJETOS, EL SEGUIMIENTO VISUAL, EL PROCESAMIENTO DE IMAGENES Y LA AUTOMATIZACION DEL PROCESO DE DESCRIPCION DEL CONTENIDO DE LA IMAGEN. EL EXITO LOGRADO POR LAS REDES NEURONALES PROFUNDAS SE DEBE A LA EXISTENCIA DE BASES DE DATOS MUY AMPLIAS Y AL DESARROLLO DE NUEVOS GPUS NECESARIOS PARA EL PROCESAMIENTO DE DICHAS BASES._x000D_ _x000D_ EL APRENDIZAJE MULTITAREA (MTL - POR SUS SIGLAS EN INGLES) ES UNA TECNICA MUY BIEN ESTUDIADA POR LA COMUNIDAD CIENTIFICA DE APRENDIZAJE AUTOMATICO. ESTA TECNICA CONTEMPLA VARIAS TAREAS DE UNA MANERA CONJUNTA CON EL OBJETIVO DE APROVECHAR LAS SIMILITUDES QUE COMPARTEN. COMO CONSECUENCIA, SE PUEDE CONSEGUIR UNA MEJORA DEL RENDIMIENTO QUE SI SE CONSIDERARA CADA TAREA POR SEPARADO. SU APLICACION ES APTA PARA CUALQUIER PROBLEMA DONDE HAY UN NUMERO DE TAREAS RELACIONADAS QUE APRENDER Y/O CUANDO HAY UN CONJUNTO DE DATOS PEQUEÑO PARA PODER APRENDER UNA TAREA EN PARTICULAR. EL MTL FUE INSPIRADO POR LA CAPACIDAD DEL SER HUMANO EN MEJORAR EL PROCESO DE APRENDIZAJE DE UNA TAREA SI ESTA SE REALIZA DE MANERA CONJUNTA CON OTRAS TAREAS QUE TIENEN RELACION, A DIFERENCIA DE CUANDO LA MISMA TAREA SE APRENDE DE MANERA INDIVIDUAL._x000D_ _x000D_ EL OBJETIVO DE ESTA MEMORIA ES PROPONER Y ESTUDIAR UN MARCO TEORICO COMUN QUE INCLUYA LOS DOS CONCEPTOS: ¿DEEP LEARNING¿ Y EL MTL. LA INICIATIVA DE ESTE PROYECTO SE APOYA EN LA EXISTENCIA DE UNOS RESULTADOS INICIALES PROMETEDORES EN ESTA DIRECCION. NUESTRA HIPOTESIS DE TRABAJO ES: LAS REDES NEURONALES PROFUNDAS PUEDEN BENEFICIARSE DE LA ESTRATEGIA DE APRENDIZAJE MULTITAREA, ES DECIR, LAS REDES ENTRENADAS CON VARIAS TAREAS AUMENTAN SU RENDIMIENTO EN COMPARACION CON LAS REDES ENTRENADAS CON UNA SOLA TAREA. ADEMAS, EL APRENDIZAJE MULTITAREA PUEDE SER MUY UTIL CUANDO EL CONJUNTO DE DATOS PARA UNA TAREA NO SEA SUFICIENTEMENTE GRANDE. EN ESTE CASO SE COMPENSA LA AUSENCIA DE DATOS EN UNA TAREA CON LOS DATOS DE LAS OTRAS TAREAS RESTANTES. POR ESTA RAZON, NOS PLANTEAMOS ESTUDIAR EN DETALLE EL ¿DEEP MULTITASK LEARNING¿ (DMTL) COMO PARTE DE ESTE PROYECTO. NOS PROPONEMOS DEFINIR UN MARCO GENERAL PARA EL DISEÑO DE LAS ARQUITECTURAS DMTL Y UNAS ESTRATEGIAS OPTIMAS PARA SU ENTRENAMIENTO. ADEMAS, SE ESTUDIARA EL EFECTO DE CONJUNTOS DE DATOS DESCOMPENSADOS PARA APRENDER LAS REDES DMTL. EN CONCRETO, SE APLICARAN LAS REDES DMTL A VARIAS APLICACIONES INCLUYENDO DETECCION DE OBJETOS, CLASIFICACION DE ESCENAS, DETECCION DE LA 'SALIENCY', AUTOMATIZACION DEL PROCESO DE DESCRIPCION DEL CONTENIDO DE LA IMAGEN Y SEGUIMIENTO VISUAL (TEMAS DE INVESTIGACION EN LOS QUE EL GRUPO LAMP YA ESTA INVOLUCRADO PERO QUE SE CONSIDERARAN AHORA DESDE EL PUNTO DE VISTA DE UNA SOLA TAREA). (Spanish)
0 references
DEEP LEARNING (‘DEEP LEARNING’) HAS BECOME THE TECHNIQUE PAR EXCELLENCE IN THE FIELD OF AUTOMATIC LEARNING APPLIED TO COMPUTER VISION. BASED ON ITS SUCCESS IN THE CLASSIFICATION OF IMAGES, DEEP LEARNING HAS SURPASSED (IN TERMS OF PERFORMANCE) OTHER TECHNIQUES OF THE SAME SCOPE AND IS BEING USED TODAY BY THE MAJORITY OF APPLICATIONS IN VISION BY COMPUTER, INCLUDING THE DETECTION OF ‘SALIENCY’, THE DETECTION OF OBJECTS, VISUAL TRACKING, IMAGE PROCESSING AND THE AUTOMATION OF THE PROCESS OF DESCRIBING THE CONTENTS OF THE IMAGE. The EXIT achieved by the PROFUND NEURONAL networks is due to the existence of very large data bases and the development of new GPUS necessities for the process of basic data._x000D__x000D_ The multitasking apprehending (MTL — BY YOUR SIGLAS IN ENGLISH) it’s a very good technic that has been established by the SCIENTIFIC COMMUNITY OF AUTOMATIC LEARNING. THIS TECHNIQUE CONTEMPLATES SEVERAL TASKS IN A JOINT WAY WITH THE AIM OF TAKING ADVANTAGE OF THE SIMILARITIES THEY SHARE. AS A RESULT, AN IMPROVEMENT IN PERFORMANCE CAN BE ACHIEVED THAN IF EACH TASK WERE CONSIDERED SEPARATELY. YOUR APPLICATION IS SUITABLE FOR ANY PROBLEM WHERE THERE ARE A NUMBER OF RELATED TASKS TO LEARN AND/OR WHEN THERE IS A SMALL DATASET TO BE ABLE TO LEARN A PARTICULAR TASK. MTL was initiated by the capacity of the human being to improve the apprehending process of a court if it is carried out in a way that is linked to other territories that have been linked to, at the same time as the same time, an individual way._x000D_ _x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x000D__x005F the OBJECTIVE OF THIS MEMORY IS TO PROPONER AND STUDY A THEORIC FRAMEWORK AS A COMMUNITY INCLUDING THE TWO CONCEPTS: DEEP LEARNING AND MTL. THE INITIATIVE OF THIS PROJECT IS BASED ON THE EXISTENCE OF PROMISING INITIAL RESULTS IN THIS DIRECTION. OUR WORKING HYPOTHESIS IS: DEEP NEURAL NETWORKS CAN BENEFIT FROM THE MULTITASKING LEARNING STRATEGY, I.E. MULTI-TASK-TRAINED NETWORKS INCREASE THEIR PERFORMANCE COMPARED TO SINGLE-TASK-TRAINED NETWORKS. IN ADDITION, MULTITASKING LEARNING CAN BE VERY USEFUL WHEN THE DATASET FOR A TASK IS NOT LARGE ENOUGH. IN THIS CASE THE ABSENCE OF DATA IN ONE TASK IS COMPENSATED BY THE DATA OF THE OTHER REMAINING TASKS. FOR THIS REASON, WE INTEND TO STUDY IN DETAIL THE DEEP MULTITASK LEARNING (DMTL) AS PART OF THIS PROJECT. WE INTEND TO DEFINE A GENERAL FRAMEWORK FOR THE DESIGN OF DMTL ARCHITECTURES AND OPTIMAL STRATEGIES FOR THEIR TRAINING. IN ADDITION, THE EFFECT OF DECOMPENSATED DATASETS TO LEARN DMTL NETWORKS WILL BE STUDIED. SPECIFICALLY, DMTL NETWORKS WILL BE APPLIED TO SEVERAL APPLICATIONS INCLUDING OBJECT DETECTION, CLASSIFICATION OF SCENES, DETECTION OF THE ‘SALIENCY’, AUTOMATION OF THE PROCESS OF DESCRIBING THE CONTENTS OF THE IMAGE AND VISUAL MONITORING (RESEARCH THEMES IN WHICH THE LAMP GROUP IS ALREADY INVOLVED BUT WHICH WILL NOW BE CONSIDERED FROM THE POINT OF VIEW OF A SINGLE TASK). (English)
12 October 2021
0 references
Cerdanyola del Vallès
0 references
Identifiers
TIN2016-79717-R
0 references