MEMORIA: A Memory Enhancement and MOment RetrIeval Application at the LSC2024
Alexandre Gago, Bernardo Kaluza, Eva Bartolomeu, Josefa N. S. Pandeirada, Ricardo Ribeiro, António J. R. Neves · 2024
The emergence of lifelogging, recording, storing and organizing personal data, underscores the need for efficient digital systems. Lifelog Search Challenge (LSC) promotes a space for discussing the challenges in developing these systems. MEMORIA, in its third version for LSC'24, stands out mainly for its renewed user interface, which aims for more intuitive and accessible navigation, offering a better user experience. Adding more layers and sub-layers has improved event segmentation based on hierarchical events. Another new feature is searching using an image and redesigning the home page, which offers a general summary of the user's lifelogs. Event navigation has been simplified, making navigating on a single page possible. Enhanced personalized search capabilities are made possible by a new labels page that offers a consolidated view of the system's annotations. Also, the image retrieval page has been updated with filters and informative pop-ups. Using Large Language Models (LLMs) to annotate images and events is a significant development in progress that could provide richer and more contextual descriptions. An event classification feature was created to retrieve pertinent memories automatically. The new album feature allows users to organize and explore image collections, while event search will enable users to find more specific moments using text queries. These new features are expected to make MEMORIA a relevant tool in lifelogging, with innovative solutions to the complex challenge of capturing, organizing, and accessing digital memories.