creative-graphic-design/layout-alignment

Layout Alignment

Description

The Layout Alignment metric evaluates how well layout elements are aligned with each other. This metric implements alignment scoring methods from multiple research papers, providing a comprehensive assessment of spatial organization and visual harmony in graphic layouts.

What It Measures

This metric computes alignment scores that quantify how elements in a layout adhere to alignment principles:

• —Edge alignment: How well element edges (left, right, top, bottom, center) align with each other
• —Spatial relationships: Detection of common alignment patterns (grids, columns, rows)
• —Visual coherence: Overall harmony created by consistent element positioning

Well-aligned layouts typically score lower (less alignment violation) and appear more professional and organized.

Metric Details

Implements alignment metrics from multiple influential layout generation papers:

• —NDN-Net (Lee et al., ECCV 2020): Neural Design Network alignment evaluation
• —AC-GAN (Li et al., TVCG 2021): Attribute-Conditioned GAN alignment metrics
• —CGL (Kikuchi et al., ACM MM 2021): Constrained Graphic Layout alignment scores

The metric analyzes element positioning to detect alignment relationships and violations.

Usage

Installation

pip install evaluate

Basic Example

import evaluate
import numpy as np

# Load the metric
metric = evaluate.load("creative-graphic-design/layout-alignment")

# Single layout processing
model_max_length, num_coordinates = 25, 4
bbox = np.random.rand(model_max_length, num_coordinates)
mask = np.random.choice(a=[True, False], size=(model_max_length,))
metric.add(bbox=bbox, mask=mask)
print(metric.compute())

Batch Processing Example

import evaluate
import numpy as np

# Load the metric
metric = evaluate.load("creative-graphic-design/layout-alignment")

# Batch processing
batch_size, model_max_length, num_coordinates = 512, 25, 4
batch_bbox = np.random.rand(batch_size, model_max_length, num_coordinates)
batch_mask = np.random.choice(a=[True, False], size=(batch_size, model_max_length))
metric.add_batch(bbox=batch_bbox, mask=batch_mask)
print(metric.compute())

Parameters

Initialization Parameters

This metric does not require any initialization parameters.

Computation Parameters

• —bbox (list of lists of int): Bounding boxes for elements in normalized coordinates
• —mask (list of lists of bool): Boolean mask indicating valid elements (True) vs padding (False)

Note: The mask parameter is crucial for handling variable-length layouts, where padding elements should be excluded from computation.

Returns

Returns a dictionary containing multiple alignment scores from different methods:

• —Different variants measuring alignment quality from various perspectives
• —Specific score names depend on the implementation details from referenced papers

Interpretation

• —Lower values generally indicate better alignment (fewer alignment violations)
• —Value interpretation depends on specific score variant:
• —Some scores measure alignment violations (lower is better)
• —Others measure alignment quality (higher is better)
• —Typical use: Compare relative scores between different layout generation methods

Key Insights

• —Professional layouts tend to have good alignment scores due to consistent spatial relationships
• —Grid-based layouts typically achieve better alignment than freeform designs
• —Alignment patterns (left-aligned, centered, etc.) are important for visual hierarchy

Citations

@inproceedings{lee2020neural,
  title={Neural design network: Graphic layout generation with constraints},
  author={Lee, Hsin-Ying and Jiang, Lu and Essa, Irfan and Le, Phuong B and Gong, Haifeng and Yang, Ming-Hsuan and Yang, Weilong},
  booktitle={Computer Vision--ECCV 2020: 16th European Conference, Glasgow, UK, August 23--28, 2020, Proceedings, Part III 16},
  pages={491--506},
  year={2020},
  organization={Springer}
}

@article{li2020attribute,
  title={Attribute-conditioned layout gan for automatic graphic design},
  author={Li, Jianan and Yang, Jimei and Zhang, Jianming and Liu, Chang and Wang, Christina and Xu, Tingfa},
  journal={IEEE Transactions on Visualization and Computer Graphics},
  volume={27},
  number={10},
  pages={4039--4048},
  year={2020},
  publisher={IEEE}
}

@inproceedings{kikuchi2021constrained,
  title={Constrained graphic layout generation via latent optimization},
  author={Kikuchi, Kotaro and Simo-Serra, Edgar and Otani, Mayu and Yamaguchi, Kota},
  booktitle={Proceedings of the 29th ACM International Conference on Multimedia},
  pages={88--96},
  year={2021}
}

References

• —Paper: Neural Design Network (Lee et al., ECCV 2020)
• —Paper: Attribute-Conditioned Layout GAN (Li et al., TVCG 2021)
• —Paper: Constrained Graphic Layout Generation (Kikuchi et al., ACM MM 2021)
• —Hugging Face Space: creative-graphic-design/layout-alignment

Related Metrics

• —Layout Non-Alignment: Measures spatial non-alignment between elements
• —Layout Overlap: Evaluates element overlap and spacing
• —Layout Validity: Checks basic layout validity constraints