Skip to content

πŸŽ“ Decompose Korean Component By Using Opencv

Notifications You must be signed in to change notification settings

92berra/Decompose

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 

History

34 Commits
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

ν•œκΈ€ 생성 λͺ¨λΈμ˜ 정확도 ν–₯상을 μœ„ν•œ κ΅¬μ„±μš”μ†Œ 뢄리



Introduction

ν•œκΈ€μ€ μ΄ˆμ„± 19개, 쀑성 21개, μ’…μ„± 28개(μ—†μŒ 포함)둜 μ‘°ν•©ν•˜μ—¬ λ§Œλ“€ 수 μžˆλŠ” 총 κΈ€μž μˆ˜λŠ” 11,172이닀. μ΄λŠ” ν•œκΈ€ 폰트 λ””μžμ΄λ„ˆκ°€ 직접 μž‘μ—…ν•  경우 λ§Žμ€ μ‹œκ°„κ³Ό λΉ„μš©μ΄ μ†Œμš”λ˜λ―€λ‘œ λ”₯λŸ¬λ‹μ„ 톡해 ν•΄κ²°ν•˜κ³ μž ν•œλ‹€. λΉ λ₯Έ μ†λ„λ‘œ λ°œμ „ν•˜λŠ” 기술둜 인해 μ„±λŠ₯κ³Ό 정확도가 ν–₯μƒλ˜μ—ˆμœΌλ‚˜ κ·Έ ν•œκ³„μ μ€ μ—¬μ „νžˆ μ‘΄μž¬ν•œλ‹€. λ”°λΌμ„œ λ³Έ 논문은 ν•œκΈ€ κ΅¬μ„±μš”μ†Œμ˜ 쑰합성에 쀑점을 λ‘” μ‹€ν—˜μ„ μ„±κ³΅μ μœΌλ‘œ λ§ˆμ³€μŒμ— 따라 μœ„μΉ˜ 정보λ₯Ό μ΄μš©ν•œ 데이터셋을 톡해 ν•œκΈ€ 생성 λͺ¨λΈμ˜ μ„±λŠ₯ ν–₯상을 μœ„ν•œ μ€‘μš”ν•œ λ°©ν–₯을 μ œκ³΅ν•˜λ©° ν–₯ν›„ ν•œκΈ€ 생성 연ꡬ에 κΈ°μ—¬ν•  κ²ƒμœΌλ‘œ κΈ°λŒ€ν•œλ‹€.


The total number of letters that can be created in Hangul by combining 19 initial consonants, 21 middle consonants, and 28 final consonants (including none) is 11,172. This is something that we want to solve through deep learning, as it takes a lot of time and money for Korean font designers to work on their own. Performance and accuracy have improved due to rapidly developing technology, but limitations still exist. Therefore, as this paper successfully completed an experiment focusing on the combinability of Hangul components, it provides important directions for improving the performance of Hangul generation models through a dataset using location information and is expected to contribute to future Hangul generation research do.


1) Overview


2) Korean Components


3) Korean Combination Type



Prerequisites

πŸ› οΈ In Progress: Modify framework from Tensorflow to PyTorch

  • Ubuntu 22.04.3 LTS
  • NVIDIA GeForce RTX 2080 Ti
  • Python 3.9.13
  • Tensorflow-gpu 1.15
conda create --name decompose python=3.9.13
conda activate decompose
pip install -r requirements.txt


Datasets

1) Generate Korean Font Images

# change directory to datasets

# generate content images
python datasets/font2img.py --label_file datasets/characters/50characters.txt --font_dir datasets/fonts/source --output_dir datasets/images/source

# generate target images
python datasets/font2img.py --label_file datasets/characters/50characters.txt --font_dir datasets/fonts/target --output_dir datasets/images/target --start_idx 1

2) Separate Components

python datasets/separator/separator-1type.py
python datasets/separator/separator-2type.py
python datasets/separator/separator-3type.py
python datasets/separator/separator-4type.py
python datasets/separator/separator-5type.py
python datasets/separator/separator-6type.py

3) Combine Separated Components for Training

python datasets/combine.py

4) Modify filename to sequential number for train

python datasets/name-modify.py

5) Transfer from images to tfrecords

python datasets/img2tfrecord.py 


Train

python main.py --mode train --output_dir trained_model --max_epochs 500


Test(Generate)

python main.py --mode test --output_dir result --checkpoint trained_model


Result


1) Generated result sample



2) Values of Loss, SSIM, FID
πŸ› οΈ In progress...


3) Figure: Loss comparison (MXFont, CKFont, My research)
πŸ› οΈ In progress...






Copyright. 92berra 2024

About

πŸŽ“ Decompose Korean Component By Using Opencv

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages