VISIONE: Redesigning an interactive retrieval system for lifelog search

Finding specific moments in lifelog collections is often difficult, as users must navigate large streams of temporally ordered images while refining queries on the fly. In this paper, we present a redesigned version of VISIONE, an interactive retrieval system tailored to the Lifelog Search Challenge (LSC). Building on our previous participation, the system has been restructured to better support time-constrained search scenarios and iterative query refinement. The new version introduces a modular architecture that integrates multiple retrieval modalities, including multimodal embeddings and metadata-based filtering, within a unified framework. On the interaction side, the interface has been redesigned to simplify query formulation through sequence-based temporal queries, while still allowing expert users to control retrieval strategies and employed feature representations. Additional components, such as relevance feedback and LLM-assisted query processing, further support both exploratory and task-driven search. Taken together, these design choices aim to make the system more effective in practice, enabling users to quickly narrow down large lifelog collections and converge toward relevant moments under realistic LSC conditions.