ufdatastudio/mixtec-zouche-nuttall-british-museum
Mixtec Zouche-Nuttall Labeled Dataset Introduction A Look at Figure Segments File Formats and Figure Names An Overview of Name Segments Overfiew of File Format and Naming Conventions An Overview of Scene Segments File Formats and Scene Names Missing Pages in the Video Narration Scenes that Span Multiple Pages Very Large Scenes: Scene 187 Introduction This dataset contains 3 different directories, figure-cutouts/, name-date-cutouts/, and scene-cutouts/.… See the full description on the dataset page: https://huggingface.co/datasets/ufdatastudio/mixtec-zouche-nuttall-british-museum.
Mixtec Zouche-Nuttall Labeled Dataset
- Introduction
- A Look at Figure Segments
- File Formats and Figure Names
- An Overview of Name Segments
- Overfiew of File Format and Naming Conventions
- An Overview of Scene Segments
- File Formats and Scene Names
- Missing Pages in the Video Narration
- Scenes that Span Multiple Pages
- Very Large Scenes: Scene 187
Introduction
This dataset contains 3 different directories, figure-cutouts/, name-date-cutouts/, and scene-cutouts/. Each directory contains images taken from the Mixtec Codex Zouche-Nuttall. The folders comprise images of human Mixtec figures, the names of such figures, and scenes from the Codex, respectively.
The images in all 3 of the directories discussed above are taken from the UF Data Studio Zouche-Nuttall Codex sample. Images of the orignal codex were provided as a PDF to the UF Data Studio by the British Museum. To learn more about the Mixtec, read the blog post here. In addition to covering the Semasiographic Writing System of the Mixtec civilization, the post contains more general information about who the Mixtec were, as well as their contributions.
A Look at Figure Segments
Figure segments can all be found in the figure-cutouts/ directory. These images of human Mixtec figures were extracted from pages of the Codex Zouche-Nuttall using Segment Anything by Meta AI.
File Formats and Figure Names
Below is an example of a human figure taken from the Codex.
Example 1 - A Human Figure Taken from the Codex Nuttall
With in the figure-cutouts/ directory, this file is name 086-a-004.png. The numerical values in the naming convention for each of the files denote pageNumber-figureQuality-figureNumber. The first field, pageNumber, indicates on which page in the Codex the figure was extracted from or where it can be found when referecing the full document.
The figureQuality field is used as a marker of quality for the extracted image. In this convention, images can be a, b, or c with a being a high quality image and c being an image that is not easy to pick out. A thorough explanation for how the quality marker was assessed can be found on the README file in this GitHub repository.
The figureNumber field indicates what number in chronological order the figure is on the page from which it was extracted using conventions for reading a Mixtec Codex. These conventions are discussed in detail in The Complete Codex Zouche-Nuttall: Mixtec Lineage Histories and Political Biographies by Robert Lloyd Williams.
An Overview of Name Segments
All images of Mixtec names in this folder are taken from the UF Data Studio Zouche-Nuttall Codex sample PDF. The orignal codex was provided by the British Museum and individual images of names were extracted from pages of the Codex Zouche-Nuttall using Segment Anything by Meta AI.
The source of ground truth for names came from The Complete Codex Zouche-Nuttall: Mixtec Lineage Histories and Political Biographies.
Overfiew of File Format and Naming Conventions
Below is an example of one of the images in this folder.
Example 1 - A Name Taken from the Codex Nuttall
In the folder, you will usually see a file named like so
005-004-a-017-10deer.png
For reference, this is the file name for the above image. The first 4 fields of the name denote from which scene the name comes from. The section on scene segments explains more about this. The example images is named 10 deer because it has the deer day sign and 10 beads. This is one of the days signs taken from the 260-day sacred Mixtec Calendar which is the same as the Aztec Calendar. Across the data set the images are different combinations of numbers and day signs, and each name corresponds to a specific scene. Based on this mapping, names can be attributed to figures within scenes.
An Overview of Scene Segments
All scenes in this folder are taken from the UF Data Studio Zouche-Nuttall Codex sample PDF. The orignal codex was provided by the British Museum and individual scenes were extracted using Lasso tools provided in macOS Preview.
The narration for each scene is provided in this YouTube video by the Nahuatl Channel.
File Formats and Scene Names
Below is an example of one of the images in this folder.
Example 1 - A Scene Taken from the Codex Nuttall
In the folder, you will usually see a file named like so
001-001-a-001.png
This file would represent a single scene taken from the Zouche-Nuttall Codex. The leading 3 digits are used to denote the page from which the scene comes from in the full book. The next 3 digits represent the act number and the letter immediately following this field is the quality of the image based on this rating scale. Note that act numbers and scene numbers are not the same, but this difference is explained more in-depth below. Finally, the last 3 digits are used for the number of the scene in the entire codex if it were read in chronological order according to the narration.
Scenes comprise individual images, while acts are treated almost like chapters, or collections of scenes. A more visual representation of this can be seen below.
Example 2 - A Directory Structure with Several Scenes (an Act)
In the image above, a section of act 2 and all the images from act 1 are highlighted to show how multiple scenes make up each act.
Missing Pages in the Video Narration
From what can be seen in the video narration, the Codex sample being referenced is the same as the one that was photographed by the British Museum. For better or worse, some pages are not narrated despite being included in the sample used to record the video. Below is a quick list of the pages that are left out in the narration. Note that there are no scenes in the dataset from page 8, but this is the only page that has this discrepancy.
Scenes that Span Multiple Pages
In the video there are also scenes that occur across folds of a page since the original Codex was intended to be read as one long page. Below is a list of all scenes that span two pages.
Very Large Scenes: Scene 187
On page 59, the events of a large conference held by Lord 8 Deer Jaguar Claw, a Mixtec Lord, are illustrated with the names of each of those who attended. The conference featured over 100 Mixtec Lords, and all of the events are pictured across 14 pages. This single scene is the largest in the Codex.
Example 3 - The Beginning of Lord Eight Deer's Conference