LifeSeeker: an interactive concept-based retrieval system for lifelog data

Area of Science:

• Computer Science
• Human-Computer Interaction
• Information Retrieval

Background:

• Lifelogging involves passively capturing personal daily events using wearable devices.
• This process creates a digital diary for future recollection or sharing.
• Existing lifelogging systems often lack robust search functionalities.

Purpose of the Study:

• To present LifeSeeker, a lifelog image retrieval system.
• To enhance the searchability of lifelogged data for users.
• To evaluate the system's retrieval performance with different user expertise levels.

Main Methods:

• LifeSeeker utilizes a combination of textual descriptions, spatial relationships, location information, and image similarities for retrieval.
• The system has participated in the Lifelog Search Challenge (LSC) for three consecutive years.
• An additional experiment was conducted to assess retrieval effectiveness with expert and novice users.

Main Results:

• The system demonstrated effective retrieval capabilities in the Lifelog Search Challenge.
• User studies indicated that LifeSeeker is effective for both expert and novice users.
• The interaction effectiveness was particularly noted when involving non-expert users.

Conclusions:

• LifeSeeker provides a powerful tool for retrieving specific moments from lifelogged data.
• The system's multimodal search approach enhances user experience and memory recall.
• Further research can explore optimizing user interaction for broader accessibility.