CASTLE 2024

From the catalogue. Listed by one researcher from one source page.

• id: castle-2024
• kind: egocentric
• robot or device: 10 participant point-of-view cameras plus 5 fixed exocentric cameras (class: human egocentric)
• how collected: volunteers recorded over four days in one fixed location, ego and exo views
• size as stated: 10 egocentric and 5 exocentric perspectives, four days, UHD at 50 fps; 600 hours
• hours: 600 | hours only claimed: none | episodes: not stated
• year: 2025
• organisation: Luca Rossetto, Cathal Gurrin et al.
• licence: CC BY-NC-SA 4.0 (Hugging Face tag) (class: non commercial)
• access: open download (Hugging Face CASTLE-Dataset/CASTLE2024) (class: open)
• link (dataset): https://huggingface.co/datasets/CASTLE-Dataset/CASTLE2024
• Hugging Face repository: CASTLE-Dataset/CASTLE2024, 7,979 downloads in 30 days, https://huggingface.co/datasets/CASTLE-Dataset/CASTLE2024
• note: The 600 hours sums all 15 camera streams, ego and exo together.

Confirm the licence at the link before relying on it.