Publicação
Memory retrieval through lifelogging using the MEMORIA application
| Resumo: | Lifelogging is the process of collecting, storing and analyzing personal multimodal data in digital format. The dissertation is part of the MEMORIA lifelogging system, a system already under development, designed to retrieve and analyze egocentric images, with the aim of improving the user’s ability to recall past memories. The aim of the dissertation is to implement algorithms to extract and organize information from images and videos for the analysis, automatic retrieval of possible important moments of the user, without the need for searches. A data collection was carried out in order to support the implementation of the algorithm. Data collection consisted of giving MEMORIA to participants who uploaded their personal images and rated the importance of events generated by the system with the uploaded images. The annotations assigned to the events by the system were collected with an associated importance. The first stage was to adapt MEMORIA for data collection. The system was divided into MEMORIA Light and MEMORIA Server. Light is a lighter version, without the image processing models, and Server is an external service, which includes an API for image processing. Participants therefore used MEMORIA Light on their computers, without demanding high computing requirements, with the processing done on Server via the API. The original system already allowed images to be uploaded and processed, generating various annotations of different types and subjects. It also grouped related images through event segmentation. These functionalities were adapted to support the new architecture. In addition, the assignment of annotations to events and a new feature, the classification of events by the user, were implemented. This allows annotations and their importance to the participant to be collected for each participant. Functionalities for visualizing the user’s important annotations were developed, in order to better analyze what is important to the user. The results collected showed the diversity of annotations and the different importance attributed from person to person, highlighting the variability of important moments for each person. The topic covered goes beyond existing knowledge of lifelog tools. It is hoped that it will improve memory by providing truly meaningful moments tailored to individual preferences. |
|---|---|
| Autores principais: | Bartolomeu, Eva Pomposo |
| Assunto: | Lifeloging Episodic memory Autobiographical memory Image retrieval Natural language processing Latent semantic analysis |
| Ano: | 2024 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade de Aveiro |
| Idioma: | inglês |
| Origem: | RIA - Repositório Institucional da Universidade de Aveiro |
| Resumo: | Lifelogging is the process of collecting, storing and analyzing personal multimodal data in digital format. The dissertation is part of the MEMORIA lifelogging system, a system already under development, designed to retrieve and analyze egocentric images, with the aim of improving the user’s ability to recall past memories. The aim of the dissertation is to implement algorithms to extract and organize information from images and videos for the analysis, automatic retrieval of possible important moments of the user, without the need for searches. A data collection was carried out in order to support the implementation of the algorithm. Data collection consisted of giving MEMORIA to participants who uploaded their personal images and rated the importance of events generated by the system with the uploaded images. The annotations assigned to the events by the system were collected with an associated importance. The first stage was to adapt MEMORIA for data collection. The system was divided into MEMORIA Light and MEMORIA Server. Light is a lighter version, without the image processing models, and Server is an external service, which includes an API for image processing. Participants therefore used MEMORIA Light on their computers, without demanding high computing requirements, with the processing done on Server via the API. The original system already allowed images to be uploaded and processed, generating various annotations of different types and subjects. It also grouped related images through event segmentation. These functionalities were adapted to support the new architecture. In addition, the assignment of annotations to events and a new feature, the classification of events by the user, were implemented. This allows annotations and their importance to the participant to be collected for each participant. Functionalities for visualizing the user’s important annotations were developed, in order to better analyze what is important to the user. The results collected showed the diversity of annotations and the different importance attributed from person to person, highlighting the variability of important moments for each person. The topic covered goes beyond existing knowledge of lifelog tools. It is hoped that it will improve memory by providing truly meaningful moments tailored to individual preferences. |
|---|
Atividades financiadas
Não foram encontrados projetos financiados.