2020
DOI: 10.1007/978-3-030-58580-8_29
|View full text |Cite
|
Sign up to set email alerts
|

Neural Design Network: Graphic Layout Generation with Constraints

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
4
1

Citation Types

4
75
0

Year Published

2020
2020
2023
2023

Publication Types

Select...
5
1

Relationship

1
5

Authors

Journals

citations
Cited by 89 publications
(79 citation statements)
references
References 19 publications
4
75
0
Order By: Relevance
“…In [9], authors propose an alignment evaluation on two public datasets but while our understanding of their alignment evaluation method has been confirmed by the authors, our results differ from theirs by orders of magnitude when we apply this method on the ground truth layouts of both evaluated datasets. Aside from the alignment evaluation method, [9] proposes an evaluation based on the Fréchet Inception Distance (FID).…”
Section: Deficiencies In Existing Quantitative Evaluation Methodssupporting
confidence: 55%
See 4 more Smart Citations
“…In [9], authors propose an alignment evaluation on two public datasets but while our understanding of their alignment evaluation method has been confirmed by the authors, our results differ from theirs by orders of magnitude when we apply this method on the ground truth layouts of both evaluated datasets. Aside from the alignment evaluation method, [9] proposes an evaluation based on the Fréchet Inception Distance (FID).…”
Section: Deficiencies In Existing Quantitative Evaluation Methodssupporting
confidence: 55%
“…In [9], authors propose an alignment evaluation on two public datasets but while our understanding of their alignment evaluation method has been confirmed by the authors, our results differ from theirs by orders of magnitude when we apply this method on the ground truth layouts of both evaluated datasets. Aside from the alignment evaluation method, [9] proposes an evaluation based on the Fréchet Inception Distance (FID). As described in their works, this method relies on the feature distribution of the penultimate layer of the discriminator : the distribution obtained on real layouts is compared with the distribution obtained on generated layouts and produces a score ranging from zero to positive infinity, with lower values indicating better performances.…”
Section: Deficiencies In Existing Quantitative Evaluation Methodssupporting
confidence: 55%
See 3 more Smart Citations