File size: 9,026 Bytes
34393a9
614d2cf
 
 
 
 
 
 
 
 
 
 
 
 
 
bb94006
614d2cf
 
 
 
549f3f3
1605397
 
 
549f3f3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
34393a9
614d2cf
4e4b9e6
614d2cf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
fec2554
 
 
614d2cf
153fbdd
614d2cf
 
 
9de292c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b79c084
614d2cf
 
 
9de292c
 
 
 
 
 
 
 
614d2cf
 
 
9de292c
 
614d2cf
 
 
9de292c
 
614d2cf
 
9de292c
 
 
 
 
 
614d2cf
 
 
9de292c
 
 
 
614d2cf
 
 
9de292c
614d2cf
 
 
9de292c
 
 
 
614d2cf
 
 
 
 
 
 
 
9de292c
 
df8cb61
9de292c
614d2cf
 
 
 
 
 
 
 
 
 
df8cb61
614d2cf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
df8cb61
9de292c
df8cb61
9de292c
 
66d5377
9de292c
614d2cf
 
 
 
 
 
 
 
 
 
9de292c
614d2cf
 
9de292c
614d2cf
 
 
9de292c
 
 
 
 
 
 
 
 
614d2cf
 
 
1605397
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
---
annotations_creators:
- expert-generated
language_creators:
- expert-generated
license:
- cc0-1.0
multilinguality:
- other-iconclass-metadata
size_categories:
- 10K<n<100K
source_datasets: []
task_categories:
- image-classification
- image-to-text
- feature-extraction
task_ids:
- multi-class-image-classification
- multi-label-image-classification
- image-captioning
pretty_name: 'Brill Iconclass AI Test Set '
tags:
- lam
- art
dataset_info:
  features:
  - name: image
    dtype: image
  - name: label
    list: string
  splits:
  - name: train
    num_bytes: 3281967920.848
    num_examples: 87744
  download_size: 3313602175
  dataset_size: 3281967920.848
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

# Dataset Card for Brill Iconclass AI Test Set

## 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://iconclass.org/testset/](https://iconclass.org/testset/)
- **Repository:**[https://iconclass.org/testset/](https://iconclass.org/testset/)
- **Paper:**[https://iconclass.org/testset/ICONCLASS_and_AI.pdf](https://iconclass.org/testset/ICONCLASS_and_AI.pdf)
- **Leaderboard:**
- **Point of Contact:**[info@iconclass.org](mailto:info@iconclass.org)

### Dataset Summary

> A test dataset and challenge to apply machine learning to collections described with the Iconclass classification system.

This dataset contains `87749` images with [Iconclass](https://iconclass.org/) metadata assigned to the images. The [iconclass](https://iconclass.org/) metadata classification system is intended to provide ['the comprehensive classification system for the content of images.'](https://iconclass.org/).

> Iconclass was developed in the Netherlands as a standard classification for recording collections, with the idea of assembling huge databases that will allow the retrieval of images featuring particular details, subjects or other common factors. It was developed in the 1970s and was loosely based on the Dewey Decimal System because it was meant to be used in art library card catalogs. [source](https://en.wikipedia.org/wiki/Iconclass)

The [Iconclass](https://iconclass.org) 

> view of the world is subdivided in 10 main categories...An Iconclass concept consists of an alphanumeric class number (“notation”) and a corresponding content definition (“textual correlate”). An object can be tagged with as many concepts as the user sees fit. [source](https://iconclass.org/)

These ten divisions are as follows:

- 0 Abstract, Non-representational Art
- 1 Religion and Magic
- 2 Nature
- 3 Human being, Man in general
- 4 Society, Civilization, Culture
- 5 Abstract Ideas and Concepts
- 6 History
- 7 Bible
- 8 Literature
- 9 Classical Mythology and Ancient History

Within each of these divisions further subdivision's are possible (9 or 10 subdivisions). For example, under `4 Society, Civilization, Culture`, one can find: 

- 41 · material aspects of daily life
- 42 · family, descendance
- 43 · recreation, amusement
- 44 · state; law; political life
- ... 

See [https://iconclass.org/4](https://iconclass.org/4) for the full list. 


To illustrate we can look at some example Iconclass classifications. 

`41A12` represents `castle`. This classification is generated via building from the 'base' division `4`, with the following attributes: 

- 4 · Society, Civilization, Culture
- 41 · material aspects of daily life
- 41A · housing
- 41A1 · civic architecture; edifices; dwellings 

[source](https://iconclass.org/41A12)

The construction of Iconclass of parts makes it particularly interesting (and challenging) to tackle via Machine Learning. Whilst one could tackle this dataset as a (multi) label image classification problem, this is only one way of tackling it. For example in the above label `castle` giving the model the 'freedom' to predict only a partial label could result in the prediction `41A` i.e. housing. Whilst a very particular form of housing this prediction for 'castle' is not 'wrong' so much as it is not as precise as a human cataloguer may provide. 

### Supported Tasks and Leaderboards

As discussed above this dataset could be tackled in various ways:

- as an image classification task
- as a multi-label classification task 
- as an image to text task
- as a task whereby a model predicts partial sequences of the label. 

This list is not exhaustive. 

### Languages

This dataset doesn't have a natural language. The labels themselves can be treated as a form of language i.e. the label can be thought of as a sequence of tokens that construct a 'sentence'. 


## Dataset Structure

The dataset contains a single configuration. 

### Data Instances

An example instance of the dataset is as follows: 

``` python
{'image': <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=390x500 at 0x7FC7FFBBD2D0>,
 'label': ['31A235', '31A24(+1)', '61B(+54)', '61B:31A2212(+1)', '61B:31D14']}
```

### Data Fields

The dataset is made up of

- an image 
- a sequence of Iconclass labels 

### Data Splits

The dataset doesn't provide any predefined train, validation or test splits. 

## Dataset Creation

> To facilitate the creation of better models in the cultural heritage domain, and promote the research on tools and techniques using Iconclass, we are making this dataset freely available. All that we ask is that any use is acknowledged and results be shared so that we can all benefit. The content is sampled from the Arkyves database. [source](https://labs.brill.com/ictestset/) 

[More Information Needed]

### Curation Rationale

[More Information Needed]

### Source Data

#### Initial Data Collection and Normalization

The images are samples from the [Arkyves database](https://brill.com/view/db/arko?language=en). This collection includes images from 

> from libraries and museums in many countries, including the Rijksmuseum in Amsterdam, the Netherlands Institute for Art History (RKD), the Herzog August Bibliothek in Wolfenbüttel, and the university libraries of Milan, Utrecht and Glasgow. [source](https://brill.com/view/db/arko?language=en)

[More Information Needed]

#### Who are the source language producers?

[More Information Needed]

### Annotations

#### Annotation process

The annotations are derived from the source dataset see above. Most annotations were likely created by staff with experience with the Iconclass metadata schema. 

#### Who are the annotators?

[More Information Needed]

### Personal and Sensitive Information

[More Information Needed]

## Considerations for Using the Data

### Social Impact of Dataset

[More Information Needed]

### Discussion of Biases

Iconclass as a metadata standard absorbs biases from the time and place of its creation (1940s Netherlands). In particular, '32B human races, peoples; nationalities' has been subject to criticism. '32B36 'primitive', 'pre-modern' peoples' is one example of a category which we may not wish to adopt. In general, there are components of the subdivisions of `32B` which reflect a belief that race is a scientific category rather than socially constructed. 

The Iconclass community is actively exploring these limitations; for example, see [Revising Iconclass section 32B human races, peoples; nationalities](https://web.archive.org/web/20210425131753/https://iconclass.org/Updating32B.pdf). 


One should be aware of these limitations to Iconclass, and in particular, before deploying a model trained on this data in any production settings. 

[More Information Needed]

### Other Known Limitations

[More Information Needed]

## Additional Information

### Dataset Curators

Etienne Posthumus

### Licensing Information
[CC0 1.0](https://creativecommons.org/publicdomain/zero/1.0/)

### Citation Information

```
@MISC{iconclass,
title = {Brill Iconclass AI Test Set},
author={Etienne Posthumus},
year={2020}
}

```


### Contributions

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