Four-Day Lifelogging Dataset with Aligned Sensor and Video Feeds

CASTLE2024 provides time-aligned sensor and video data captured from 10 participants over four days in a controlled environment. The dataset, created by the CASTLE-Dataset organization, is designed for research in lifelogging and multimodal retrieval. It was last updated on the Hugging Face platform in February 2026.

Use Cases

• Training models for human activity recognition using synchronized video and sensor streams.
• Benchmarking multimodal retrieval systems on aligned sensor and video data pairs.
• Studying temporal patterns in lifelogging data across a four-day observation period.

Strengths

• Data from 10 distinct participants provides multiple subject perspectives.
• Time-aligned multimodal data enables joint analysis of sensor and video streams.

Limitations

• Small participant pool of 10 individuals limits demographic diversity.
• Data captured over only four days may not capture long-term activity patterns.
• Collection in a controlled environment may not reflect real-world conditions.

Provenance

Source

CASTLE-Dataset organization via Hugging Face.

Collection Method

Captured in a controlled environment from 10 participants.

Time Range

Four days of observation.

Freshness

Last updated on Hugging Face in February 2026.