The authors trained NAMVIS to generate images of unseen objects from new camera positions, given one or more source images and their known camera poses.
Autoregressive synthesis generates novel object views faster than diffusion
NAMVIS predicts geometry-conditioned visual tokens from coarse to fine, improving benchmark fidelity while cutting inference time.
Top University
Ramil Khafizov · Ilya Statsenko · Ruslan Rakhimov · Artem Komarichev · Peter Wonka · Evgeny Burnaev
Applied AI Institute · T-Tech · KAUST · AXXX
Research Digest··3 min read
Khafizov et al.
Why this paper
From KAUST and 3 others
In one line
NAMVIS predicts novel views via coarse-to-fine autoregression, outperforming diffusion models in quality and speed.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors (2 noted)
- ✓Reports numbers on named benchmarks
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§