Parameter-Efficient Fine-Tuning of DINOv2 for Large-Scale Font Classification
Fuente:
arXiv
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915911309459456 |
|---|---|
| author | Chen, Daniel Zinn, Zaria Lowe, Marcus |
| author_facet | Chen, Daniel Zinn, Zaria Lowe, Marcus |
| contents | We introduce GoogleFontsBench, the first public benchmark for classifying open-source web fonts, addressing a gap left by existing benchmarks that cover only commercial typefaces. GoogleFontsBench comprises 394 font variants across 32 Google Fonts families, a reproducible synthetic data generation pipeline (~575 images per variant, ~226K total), and a typographically-grounded evaluation metric (SWER) that weights errors by visual severity. We establish baselines using six fine-tuning strategies on a DINOv2 Vision Transformer backbone. Parameter-efficient adaptation with LoRA achieves 99.0% top-1 accuracy while training only 1% of the model's 87.2M parameters, with errors 140x less severe than random guessing. We release the benchmark, all trained models, and the full training pipeline as open-source resources. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_13889 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Parameter-Efficient Fine-Tuning of DINOv2 for Large-Scale Font Classification Chen, Daniel Zinn, Zaria Lowe, Marcus Computer Vision and Pattern Recognition Machine Learning We introduce GoogleFontsBench, the first public benchmark for classifying open-source web fonts, addressing a gap left by existing benchmarks that cover only commercial typefaces. GoogleFontsBench comprises 394 font variants across 32 Google Fonts families, a reproducible synthetic data generation pipeline (~575 images per variant, ~226K total), and a typographically-grounded evaluation metric (SWER) that weights errors by visual severity. We establish baselines using six fine-tuning strategies on a DINOv2 Vision Transformer backbone. Parameter-efficient adaptation with LoRA achieves 99.0% top-1 accuracy while training only 1% of the model's 87.2M parameters, with errors 140x less severe than random guessing. We release the benchmark, all trained models, and the full training pipeline as open-source resources. |
| title | Parameter-Efficient Fine-Tuning of DINOv2 for Large-Scale Font Classification |
| topic | Computer Vision and Pattern Recognition Machine Learning |
| url | https://arxiv.org/abs/2602.13889 |