Metraipolis › archive › dc_d143cd9aa878

From UNI2-h to ConvNeXt-T: Lightweight Nuclei Instance Segmentation via Knowledge Distillation — arXiv:2609.35203

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.35203
License (as archived)
CC0-1.0
Archived
2026-10-07T23:52:42.284Z
SHA-256 of the archived text
bc5e8893533c041908f42675f80eeee7ba5df7afe5cfdc789e4c05479becb505
Stable URL of this page
https://metraipolis.com/documento/dc_d143cd9aa878

How to cite

From UNI2-h to ConvNeXt-T: Lightweight Nuclei Instance Segmentation via Knowledge Distillation — arXiv:2609.35203 — arXiv (en), https://arxiv.org/abs/2609.35203 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d143cd9aa878 (Metraipolis archive dc_d143cd9aa878, sha256 bc5e8893533c041908f42675f80eeee7ba5df7afe5cfdc789e4c05479becb505)
BibTeX
@misc{dcd143cd9aa878,
  title = {From UNI2-h to ConvNeXt-T: Lightweight Nuclei Instance Segmentation via Knowledge Distillation — arXiv:2609.35203},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.35203},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d143cd9aa878; sha256 bc5e8893533c041908f42675f80eeee7ba5df7afe5cfdc789e4c05479becb505},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.35203. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Nuclei instance segmentation is a core task in digital pathology, yet high-accuracy models rely on large vision transformer (ViT) encoders whose inference speed cannot meet real-time clinical demands. We propose a lightweight scheme that distills the UNI2-h pathology foundation model into a ConvNeXt-Tiny student (Ours-T, 34.7M parameters, 1/20 of the teacher) via output-level knowledge distillation. Ours-T achieves an mPQ of 0.519 on PanNuke (98.8% of the teacher), a zero-shot bPQ of 0.668 on MoNuSeg, and an inference speed of 634.3 img/s, requiring only 0.045 s for full-resolution 1024^2 analysis (21.8x speedup). Experiments further show that multi-scale gated convolution (MALA) yields no gain under ViT encoders, and output-level distillation alone suffices for efficient knowledge transfer.

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-29 in cs.CV).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2609.35203 — 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 2026-10-07T23:29:27.535Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).