JungaoCanada/dl3dv-downsampled

DL3DV — Downsampled Videos (1080p, half fps)

Downsampled derivative of DL3DV/DL3DV-ALL-video (all 10,203 scene videos of the DL3DV-10K dataset).

Processing

Each original video (typically 3840×2160 @ 60 fps, HEVC ~105 Mbps) was re-encoded with:

• fps halved (e.g. 60 → 30, 59.94 → 29.97; exact rational halving)
• resolution halved: long side scaled to 1920 px, aspect ratio preserved (portrait videos scaled to height/width 1920 accordingly; videos already ≤1920 were not upscaled)
• codec: x265 (HEVC), CRF 21, preset medium, audio removed
• total size: 6.9 TB → ~1.4 TB

Layout

Identical to the source repo: <subset>/<scene_hash>/video.mp4, subsets 1K … 11K.

License / attribution

This is a derivative of the DL3DV-10K dataset and inherits its license and terms of use (CC BY-NC 4.0, research use). Please cite the original DL3DV paper and comply with the original dataset's terms.