Differentiable Hierarchical Visual Tokenization
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866914135812341760 |
|---|---|
| author | Aasan, Marius Hjelkrem-Tan, Martine Catalano, Nico Choi, Changkyu Rivera, Adín Ramírez |
| author_facet | Aasan, Marius Hjelkrem-Tan, Martine Catalano, Nico Choi, Changkyu Rivera, Adín Ramírez |
| contents | Vision Transformers rely on fixed patch tokens that ignore the spatial and semantic structure of images. In this work, we introduce an end-to-end differentiable tokenizer that adapts to image content with pixel-level granularity while remaining backward-compatible with existing architectures for retrofitting pretrained models. Our method uses hierarchical model selection with information criteria to provide competitive performance in both image-level classification and dense-prediction tasks, and even supports out-of-the-box raster-to-vector conversion. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_02652 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Differentiable Hierarchical Visual Tokenization Aasan, Marius Hjelkrem-Tan, Martine Catalano, Nico Choi, Changkyu Rivera, Adín Ramírez Computer Vision and Pattern Recognition 68T45 I.2.10; I.4.10; I.4.6; I.3.8 Vision Transformers rely on fixed patch tokens that ignore the spatial and semantic structure of images. In this work, we introduce an end-to-end differentiable tokenizer that adapts to image content with pixel-level granularity while remaining backward-compatible with existing architectures for retrofitting pretrained models. Our method uses hierarchical model selection with information criteria to provide competitive performance in both image-level classification and dense-prediction tasks, and even supports out-of-the-box raster-to-vector conversion. |
| title | Differentiable Hierarchical Visual Tokenization |
| topic | Computer Vision and Pattern Recognition 68T45 I.2.10; I.4.10; I.4.6; I.3.8 |
| url | https://arxiv.org/abs/2511.02652 |