ScaleNav Internet Navigation Videos

Source: SpatialVID-HQ, Jiahao Wang, Yufeng Yuan and collaborators, CVPR 2026.
Paper: https://arxiv.org/abs/2509.09676
Project: https://github.com/NJU-3DV/SpatialVID
Dataset: https://huggingface.co/datasets/FelixYuan-YF/SpatialVID-HQ
Source revision: 97e22f97447c012ef7460e5989b83a674643a880

The SpatialVID-derived videos and ScaleNav annotations are distributed under
Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International.
License: https://creativecommons.org/licenses/by-nc-sa/4.0/
The complete license, including its disclaimer of warranties, is in LICENSE.
Retain attribution and license notices, identify further changes, use the data
noncommercially, and distribute adaptations under the same license.

ScaleNav selected clips and navigation intervals, resampled video to 4 Hz and
512 x 288 pixels, and added navigation instructions, paraphrases,
and fixed training/validation splits. This release preserves the prepared
training MP4 files without another video encoding pass. Frame intervals are
zero-based and end-exclusive on the prepared 4 Hz timeline. Each archive's
annotations.jsonl retains upstream clip IDs, archive paths and source-video
paths. The release index additionally records video hashes and frame counts.

Copyright 2026 ScaleNav Authors applies to our original contributions, not to
the underlying footage. Original rights remain with their respective holders.
The license grants only rights that the licensors have authority to grant;
it does not confer additional privacy, publicity or other third-party rights.
This is an independently prepared subset, not an official SpatialVID release
or an endorsement by its authors. No InteriorGS or HM3D scene data is included.

For attribution corrections or removal requests, identify the affected clip IDs
at https://huggingface.co/datasets/Yangyihui/ScaleNav-Data/discussions or
https://github.com/yangyihui0502/ScaleNav/issues .