Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/mountain/metafont-glyphs
A metafont-glyphs dataset which facilitate people to define CJK-like glyphs with their metafont scripts by machine learning
https://github.com/mountain/metafont-glyphs
ancient-languages dataset glyphs machine-learning metafont
Last synced: 30 days ago
JSON representation
A metafont-glyphs dataset which facilitate people to define CJK-like glyphs with their metafont scripts by machine learning
- Host: GitHub
- URL: https://github.com/mountain/metafont-glyphs
- Owner: mountain
- License: mit
- Created: 2023-01-27T21:35:27.000Z (almost 2 years ago)
- Default Branch: main
- Last Pushed: 2024-05-22T16:32:26.000Z (7 months ago)
- Last Synced: 2024-05-22T16:37:46.114Z (7 months ago)
- Topics: ancient-languages, dataset, glyphs, machine-learning, metafont
- Language: Python
- Homepage:
- Size: 5.34 MB
- Stars: 10
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# metafont-glyphs dataset
This repository contains a dataset of glyph images generated from random metafont scripts.
We have many variant characters from ancient literature, such as oracle-bone inscriptions, bronze inscriptions etc.
The main difficulty is that these characters are not standardized, and in order to digitize them, we need to collect a large number of glyph images,
and then use machine learning to generate metafont scripts from glyph images.
It is a different problem from OCR, which is focused on recognizing and transcribing written text from images.
The goal is to develop models that can generate accurate geometric descriptions of these glyphs,
which can then be used in typesetting systems such as Metafont and Tex.
## gallery
## An example
The following glyph is generated from the following metafont script.
```metafont
% file name: bcdylw.mf
% mode_setup;
% Define a random shape for the training corpus
beginchar("bcdylw",12pt#,12pt#,0);
% Setup coordinates as an equation system
x12 = 1 * w / 10;
x14 = 1 * w / 10;
x32 = 3 * w / 10;
x34 = 3 * w / 10;
x36 = 3 * w / 10;
x38 = 3 * w / 10;
x45 = 4 * w / 10;
x68 = 6 * w / 10;
y12 = 2 * w / 10;
y14 = 4 * w / 10;
y32 = 2 * w / 10;
y34 = 4 * w / 10;
y36 = 6 * w / 10;
y38 = 8 * w / 10;
y45 = 5 * w / 10;
y68 = 8 * w / 10;
% Draw the character curve
% z1 is the same as (x1, y1)
pickup pencircle xscaled 0.06w yscaled 0.02w rotated 243;
draw z36..z38;
draw z36..z38;
draw z12..z14;
draw z45..z68..z34;
draw z32..z34;
endchar;
end
```
To facilitate machine learning, we also provide the following csv file for the sequence of coordinates of the control points.
```csv
-0.1666, +0.5750
-0.5000, +0.2497
-0.8333, +0.6750
+0.3000, +0.6000
+0.3000, +0.8000
-0.5000, -0.5000
+0.3000, +0.6000
+0.3000, +0.8000
-0.5000, -0.5000
+0.1000, +0.2000
+0.1000, +0.4000
-0.5000, -0.5000
+0.4000, +0.5000
+0.6000, +0.8000
+0.3000, +0.4000
-0.5000, -0.5000
+0.3000, +0.2000
+0.3000, +0.4000
-0.5000, -0.5000
```
Lines with negative coordinates are special
* `-0.1666, +0.5750` means the xscale of the pen is 0.0575
* `-0.5000, +0.2497` means the yscale of the pen is 0.02497
* `-0.8333, +0.6750` means the rotation of the pen is 0.675 * 360 degrees, i.e. 243 degrees
* `-0.5000, -0.5000` means the pen is lifted up, i.e. the pen is not drawing and hence the end of the curve
## How to use
### Install texlive
```bash
sudo apt install texlive-full # for Debian and Ubuntu
brew install texlive # for macOS
```
### Install python3, ninja and imagemagick
```bash
sudo apt install python3 # for Debian and Ubuntu
sudo apt install ninja-build # for Debian and Ubuntu
sudo apt install imagemagick # for Debian and Ubuntu
brew install python3 # for macOS
brew install ninja # for macOS
brew install imagemagick # for macOS
pip3 install -r requirements.txt # for any platform
```
### Fire the tests
```bash
git clone https://github.com/mountain/metafont-glyphs.git
cd metafont-glyphs
bin/meg fontg 48
bin/meg build
```
You should see the glyphs in the `data/glyph` folder.
### Make dataset
```bash
cd metafont-glyphs
bin/meg clean-all
bin/meg fontg 100000
bin/meg build
bin/meg dsmk
```
You should see three parquet files in the `data/dataset` folder.