RECORD
ArtFID: Quantitative Evaluation of Neural Style Transfer
Computer Vision and Pattern Recognition · Submitted 25 Jul 2022
abstract compiled · 2026-09-23 · no language-model review
The authors' main claim We provide extensive evaluations and a large-scale user study to show that the proposed metric strongly coincides with human judgment.
The compiled record
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
Open the compiled record
The abstract, sentence by sentence
The field of neural style transfer has experienced a surge of research exploring different avenues ranging from optimization-based approaches and feed-forward models to meta-learning methods. The developed techniques have not just progressed the field of style transfer, but also led to breakthroughs in other areas of computer vision, such as all of visual synthesis. However, whereas quantitative evaluation and benchmarking have become pillars of computer vision research, the reproducible, quantitative assessment of style transfer models is still lacking. Even in comparison to other fields of visual synthesis, where widely used metrics exist, the quantitative evaluation of style transfer is still lagging behind. To support the automatic comparison of different style transfer approaches and to study their respective strengths and weaknesses, the field would greatly benefit from a quantitative measurement of stylization performance. Therefore, we propose a method to complement the currently mostly qualitative evaluation schemes. We provide extensive evaluations and a large-scale user study to show that the proposed metric strongly coincides with human judgment.
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | The field of neural style transfer has experienced a surge of research exploring different avenues ranging from optimization-based approaches and feed-forward models to meta-learning methods. | background 1.00 | 0.04 | 0.02 | asserted 1.00 |
| s1 | The developed techniques have not just progressed the field of style transfer, but also led to breakthroughs in other areas of computer vision, such as all of visual synthesis. | background 1.00 | 0.05 | 0.04 | asserted 1.00 |
| s2 | However, whereas quantitative evaluation and benchmarking have become pillars of computer vision research, the reproducible, quantitative assessment of style transfer models is still lacking. | gap 1.00 | 0.06 | 0.05 | asserted 1.00 |
| s3 | Even in comparison to other fields of visual synthesis, where widely used metrics exist, the quantitative evaluation of style transfer is still lagging behind. | gap 0.79 | 0.05 | 0.03 | asserted 1.00 |
| s4 | To support the automatic comparison of different style transfer approaches and to study their respective strengths and weaknesses, the field would greatly benefit from a quantitative measurement of stylization performance. | gap 0.48 | 0.05 | 0.08 | hedged possibility 0.73 |
| s5 | Therefore, we propose a method to complement the currently mostly qualitative evaluation schemes. | contribution 1.00 | 0.62 | 0.13 | asserted 0.94 |
| s6 | We provide extensive evaluations and a large-scale user study to show that the proposed metric strongly coincides with human judgment. | result 0.84 | 0.96 | 0.67 | asserted 1.00 |
What was examined, and what was not
- arXiv listing abstract v1
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
The paper
Original abstract
The field of neural style transfer has experienced a surge of research exploring different avenues ranging from optimization-based approaches and feed-forward models to meta-learning methods. The developed techniques have not just progressed the field of style transfer, but also led to breakthroughs in other areas of computer vision, such as all of visual synthesis. However, whereas quantitative evaluation and benchmarking have become pillars of computer vision research, the reproducible, quantitative assessment of style transfer models is still lacking. Even in comparison to other fields of visual synthesis, where widely used metrics exist, the quantitative evaluation of style transfer is still lagging behind. To support the automatic comparison of different style transfer approaches and to study their respective strengths and weaknesses, the field would greatly benefit from a quantitative measurement of stylization performance. Therefore, we propose a method to complement the currently mostly qualitative evaluation schemes. We provide extensive evaluations and a large-scale user study to show that the proposed metric strongly coincides with human judgment.
Human review
How the literature uses it
1 arXiv paper cite this in 1 sentence
Papers citing it
Uses its method or tool 1 paper
-
For quantitative comparison, we adopt metrics such as LPIPS, FID and ArtFID [ref] .
Inversion-Free Video Style Transfer with Trajectory Reset Attention Control and Content-Style Bridging · Image Style Transfer · 0.96
What it cites · 5 sentences without a confident label
Compares with 1 paper
-
We observe that the ranking induced by the ArtFID coincides with the majority of user studies reported in the literature [ref] .
Consistent Style Transfer · Correlation with Human Judgment · 0.88
Cites as background 1 paper
-
Another line of research used attention mechanisms to integrate local style patterns and align content and style manifolds [ref] .
Consistent Style Transfer · Related Work · 0.99
Discussion
Know something about this paper? Leave a comment, or highlight any passage above to comment on it.