eezy/basic_shapes_1000

---
dataset_info:
- config_name: mixed
features:
- name: svg
dtype: string
- name: png
dtype: image
- name: layer_mask
dtype: image
- name: object_mask
dtype: image
- name: segments
dtype:
array3_d:
shape:
- -1
- 512
- 512
dtype: bool
splits:
- name: train
num_bytes: 825420657
num_examples: 3200
- name: validation
num_bytes: 103491703
num_examples: 400
- name: test
num_bytes: 14362883
num_examples: 400
download_size: 79499715
dataset_size: 943275243
- config_name: circles
features:
- name: svg
dtype: string
- name: png
dtype: image
- name: layer_mask
dtype: image
- name: object_mask
dtype: image
- name: segments
dtype:
array3_d:
shape:
- -1
- 512
- 512
dtype: bool
splits:
- name: train
num_bytes: 202172900
num_examples: 800
- name: validation
num_bytes: 25380696
num_examples: 100
- name: test
num_bytes: 3587893
num_examples: 100
download_size: 28664837
dataset_size: 231141489
- config_name: squares
features:
- name: svg
dtype: string
- name: png
dtype: image
- name: layer_mask
dtype: image
- name: object_mask
dtype: image
- name: segments
dtype:
array3_d:
shape:
- -1
- 512
- 512
dtype: bool
splits:
- name: train
num_bytes: 209226435
num_examples: 800
- name: validation
num_bytes: 26362720
num_examples: 100
- name: test
num_bytes: 3590905
num_examples: 100
download_size: 10376213
dataset_size: 239180060
- config_name: squares_and_circles
features:
- name: svg
dtype: string
- name: png
dtype: image
- name: layer_mask
dtype: image
- name: object_mask
dtype: image
- name: segments
dtype:
array3_d:
shape:
- -1
- 512
- 512
dtype: bool
splits:
- name: train
num_bytes: 207141741
num_examples: 800
- name: validation
num_bytes: 25735545
num_examples: 100
- name: test
num_bytes: 3590235
num_examples: 100
download_size: 20138547
dataset_size: 236467521
- config_name: scer
features:
- name: svg
dtype: string
- name: png
dtype: image
- name: layer_mask
dtype: image
- name: object_mask
dtype: image
- name: segments
dtype:
array3_d:
shape:
- -1
- 512
- 512
dtype: bool
splits:
- name: train
num_bytes: 206879581
num_examples: 800
- name: validation
num_bytes: 26012748
num_examples: 100
- name: test
num_bytes: 3593856
num_examples: 100
download_size: 20320118
dataset_size: 236486185
---

# Dataset Card for BasicShapes1000

## Table of Contents
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)

## Dataset Description

- **Homepage:** https://eezy.com

### Dataset Summary

This is a synthetic dataset containing randomly-generated SVGs with various shapes
### Supported Tasks and Leaderboards

NA

### Languages

NA

## Dataset Structure

The dataset is composed of 4 base domains, plus a 'mixed' domain that is a superset of the other 4:
* `circles` - only circles
* `squares` - only squares
* `squares_and_circles` - circles and squares present in the same svg
* `scer` - squares, circles, ellipses, and rectangles present in the same svg
* `mixed` - an aggregation of all of the above

### Data Instances

There's stuff there

### Data Fields

Each example has 4 fields:

* `svg` - the raw svg as a string
* `png` - a raster rendering of the svg with a white background
* `object_mask` - a black/white mask that defines the outlines of the svg objects
* `layer_mask` - a greyscale mask that defines layers of svg objects - overlap regions are brighter. Created by making all the objects white and semi-transparent
* `segments` - a numpy array in the shape `(N,512,512), dtype='bool'` where N is the number of svg objects. The array is a mask of each object with `True` in the area of the object

### Data Splits

Train & validation include the layer and object masks, test does not

## Dataset Creation

Generated by randomly inserting objects into an SVG.

### Curation Rationale

Objects should have at least 50% of their bounding box visible - i.e. no big circle completely obscuring a little circle

### Source Data

`/dev/urandom`

#### Initial Data Collection and Normalization

NA

#### Who are the source language producers?

NA

### Annotations

see [Data Fields](#data-fields)

#### Annotation process

see [Data Fields](#data-fields)

#### Who are the annotators?

Imagemagick/pysvg

### Personal and Sensitive Information

Unlikely

## Considerations for Using the Data

Please do not use for world domination.

### Social Impact of Dataset

NA

### Discussion of Biases

Dataset is highly biased against triangles and concave shapes

### Other Known Limitations

Color selection is pretty limited.

## Additional Information

### Dataset Curators

[Aleks Clark](https://github.com/aleksclark)

### Licensing Information

CC-BY

### Citation Information

Link it I guess?

### Contributions

Thanks to [@aleksclark](https://github.com/aleksclark) for adding this dataset.

原始信息汇总

数据集卡片 for BasicShapes1000

数据集描述

数据集摘要

这是一个包含随机生成的SVG文件的合成数据集，包含各种形状。

支持的任务和排行榜

NA

语言

NA

数据集结构

数据集由4个基础域和一个“混合”域组成，混合域是其他4个域的超集：

• circles - 仅包含圆形
• squares - 仅包含正方形
• squares_and_circles - 包含圆形和正方形
• scer - 包含正方形、圆形、椭圆形和矩形
• mixed - 包含上述所有形状

数据实例

NA

数据字段

每个示例包含以下字段：

• svg - 原始SVG字符串
• png - 带有白色背景的SVG光栅渲染
• object_mask - 定义SVG对象轮廓的黑白掩码
• layer_mask - 定义SVG对象层的灰度掩码 - 重叠区域更亮。通过使所有对象为白色和半透明创建
• segments - 形状为(N,512,512), dtype=bool的numpy数组，其中N是SVG对象的数量。该数组是每个对象的掩码，对象区域为True

数据分割

训练和验证集包含层和对象掩码，测试集不包含

数据集创建

通过随机插入对象到SVG中生成。

策划理由

对象应至少有50%的边界框可见 - 即没有大圆完全遮挡小圆

源数据

/dev/urandom

初始数据收集和规范化

NA

源语言生产者

NA

注释

参见数据字段

注释过程

参见数据字段

注释者

Imagemagick/pysvg

个人和敏感信息

不太可能

使用数据的注意事项

请勿用于世界统治。

数据集的社会影响

NA

偏见讨论

数据集对三角形和凹形有高度偏见

其他已知限制

颜色选择相当有限。

附加信息

数据集策展人

Aleks Clark

许可信息

CC-BY

引用信息

链接它吧

贡献

感谢@aleksclark添加此数据集。

搜集汇总

背景与挑战

背景概述

该数据集是一个合成数据集，包含1000个随机生成的SVG图形，涵盖圆形、方形、椭圆和矩形等多种形状，但不包括三角形和凹形。每个图形示例提供SVG原始字符串、PNG渲染图及多种掩码信息，适用于图形识别和处理任务。数据集采用CC-BY许可，由Aleks Clark创建。