Metraipolis › archive › dc_9770469dc4e2

PixelDense: Dense Prediction as Representation Alignment for Pixel Diffusion — arXiv:2610.00483

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.00483
License (as archived): CC0-1.0
Archived: 2026-10-06T19:11:05.030Z
SHA-256 of the archived text: 2934a3f0a4198ac235450e41b9f0befc63a782e6e05d8725decbe375c70d4790
Stable URL of this page: https://metraipolis.com/documento/dc_9770469dc4e2

How to cite

PixelDense: Dense Prediction as Representation Alignment for Pixel Diffusion — arXiv:2610.00483 — arXiv (en), https://arxiv.org/abs/2610.00483 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_9770469dc4e2 (Metraipolis archive dc_9770469dc4e2, sha256 2934a3f0a4198ac235450e41b9f0befc63a782e6e05d8725decbe375c70d4790)

BibTeX

@misc{dc9770469dc4e2,
  title = {PixelDense: Dense Prediction as Representation Alignment for Pixel Diffusion — arXiv:2610.00483},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.00483},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_9770469dc4e2; sha256 2934a3f0a4198ac235450e41b9f0befc63a782e6e05d8725decbe375c70d4790},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.00483. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Representation alignment (REPA) accelerates diffusion transformer training, but its alignment targets are almost exclusively semantic encoders such as DINOv2 and CLIP. Recent analysis points to spatial structure, not global semantics, as the carrier of the alignment effect, yet dense-prediction foundation models trained to predict that structure remain overlooked as REPA targets. In pixel-space diffusion, SAM2, Depth Anything v2, and Metric3D v2 each outperform the DINOv2-only GenEval baseline, with the two geometric teachers leading the segmentation teacher. A flat sum of all four teachers, however, lands below the best single geometric teacher, as semantic and geometric gradients compete for one denoiser projection. We introduce PixelDense, which routes DINOv2 and SAM2 through a semantic projection stream, routes Depth Anything v2 and Metric3D v2 through a geometric projection stream, and adds a weight-space orthogonality penalty that keeps the two streams in disjoint subspaces. All four teachers are frozen during training and dropped at inference. Applied to PixelGen and DeCo with a single recipe, PixelDense improves GenEval, DPG-Bench, and HPS v2.1, raises PixelGen-XXL's GenEval Overall from 0.7927 to 0.8093, and beats every single-teacher and unfactored multi-teacher variant. In partial-noise reconstruction, independent panoptic, depth, and surface-normal probes show up to 53.1% PQ gain and 36.0% depth AbsRel reduction at $\tau=0.5$ across COCO and Flickr30K. From random initialization, PixelDense also reaches the baseline's peak GenEval 1.23x faster. In SDEdit editing on PIE-Bench, PixelDense keeps more of the source background and layout at every edit strength, raising background PSNR by up to 2.2 dB.

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-10-02 in cs.CV).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2610.00483 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-06T17:37:36.669Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-10-02?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).