What question did this study set out to answer?

The aim is to improve image retrieval from lifelogs using a multi-stage approach in response to textual queries.

April 1, 2026Open Access

LifeIR at the NTCIR-18 Lifelog-6 Task

View Full Paper

Ask AI

Bookmark

Authors

Evaluation demonstrates effective image retrieval from lifelogs using a multi-stage pipeline, indicating enhanced search efficiency.

The aim is to improve image retrieval from lifelogs using a multi-stage approach in response to textual queries.
Proposed a multi-stage pipeline for lifelog image retrieval
Filtered blurred images to improve quality
Rewrote queries for clearer user intent
Extended candidate images using temporal connections
Reranked results with a multimodal large language model
Evaluation showed effectiveness of image filtering and query rewriting
Temporal connection extension enhanced candidate image set
Overall pipeline demonstrated strong relevance judgment capabilities

Chen et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark