MiraData

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

• id: miradata
• kind: world model and video
• robot or device: long-duration video with structured captions (class: other)
• how collected: curated sources with GPT-4V captions
• size as stated: HF size tag 100K<n<1M; 16,000 hours claimed, not released
• hours: not stated | hours only claimed: 16,000 | episodes: not stated
• year: 2024
• organisation: Tencent ARC
• licence: GPL-3.0 (class: open)
• access: open (class: open)
• link (dataset): https://huggingface.co/datasets/TencentARC/MiraData
• Hugging Face repository: TencentARC/MiraData, 437 downloads in 30 days, https://huggingface.co/datasets/TencentARC/MiraData
• note: Long clips with high motion strength.

Confirm the licence at the link before relying on it.