datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
scene-extrapolation-benchmark
Scene-Extrapolation Benchmark Suite
A unified, honest evaluation of generative novel-view synthesis (NVS) in the regimes where
per-scene 3D-Gaussian-Splatting reconstruction fails:
Extrapolative — held-out, unobserved views (not interpolation between dense captures).
Dynamic — moving foreground content.
Long-horizon — chained / loop-closing camera trajectories where drift accumulates.
Memory / revisit — recall when the camera returns to a previously-observed pose (do methods… See the full description on the dataset page: https://huggingface.co/datasets/luuuulinnnn/scene-extrapolation-benchmark.scene-extrapolation-dl3dv-vace14b
Scene Extrapolation: DL3DV → Wan2.1-VACE-14B (warp + inpaint)
Generate novel-view video that extends a real scene beyond the observed region. Observed frames are
depth-warped (DepthAnythingV2 + COLMAP poses) into target viewpoints → a control video with holes +
an inpaint mask → Wan2.1-VACE-14B generates the unobserved region.
Method
observed frames → forward-splat warp into target poses → [control w/ holes | mask] → VACE-14B inpaints
Data: 1,500 DL3DV scenes… See the full description on the dataset page: https://huggingface.co/datasets/luuuulinnnn/scene-extrapolation-dl3dv-vace14b.