정말 필요한 기술과 부가적인것 구분

[Wolf0605]

What you need essentialy (정말 필요한 것)

1. Text detecting, labeling

Way 1

• Using opencv, pytorch ocr

[([[19, 17], [395, 17], [395, 107], [19, 107]],
  'THE VAST WAISTBAND ',
  0.7257467526221919),
 ([[317.3183180150092, 180.29129044202583],
   [384.8298712691499, 171.78564375624916],
   [388.6816819849908, 217.70870955797417],
   [321.1701287308501, 225.21435624375084]],
  'SHLE',
  0.659963846206665)]

Way 2

• Naver clova api or Google ocr api

Naver clova : https://www.ncloud.com/product/aiService/ocr Google ocr guide : https://cloud.google.com/functions/docs/tutorials/ocr#objectives

In Naver clova you can get boundary about labels :

{
meta: {
imageSize: {
width: 1736,
height: 1492
},
domain: "general",
language: "ja"
},
words: [
{
id: 1,
boundingBox: 
[[ 647, 62 ],
[827,65],
[826,111],
[646,108]],
isVertical: false,
text: "セミナー",
confidence: 1

2. Translate

• Naver papago or Google translate

Naver papago : https://developers.naver.com/docs/papago/ Google translate : https://cloud.google.com/translate

How to use Naver papago in python : https://bziwnsizd.tistory.com/36 PaPago guide : https://guide-fin.ncloud-docs.com/docs/ocr-ocr-1-4

3. Inpainting

Example :

1. https://pythonawesome.com/internal-video-inpainting-by-implicit-long-range-propagation-in-python/ https://github.com/Tengfei-Wang/Implicit-Internal-Video-Inpainting

2. https://pythonawesome.com/aot-gan-for-high-resolution-image-inpainting/

inpainting librarys : https://pythonrepo.com/tag/inpainting

4. Rewrite

Rewrite strings in each frame

[Wolf0605]

Sub (부가적)

1. Detecting

1. If detecting wrong ex )

   事例や導入効果など具体的な話がきけて良か水

   글자가 이상 (정확도가 낮은경우) 해서 잘못 읽은 경우

RNN or LSTM 을 이용할 수 있지 않을까??

2. String color detecting 색검출 opencv ??

2. Translate

If translate wrong

3. Impainting

4. Rewrite(덮어쓰기)

1. Find similar string or color

2. 만약 글자가 바르지 않을 경우 원래 각도에 맞게 글자 돌려주기

3. 글자가 크기가 일정하지 않을 떄

[Wolf0605]