From 6449b06f89242e27ae784f3cfff749225907b156 Mon Sep 17 00:00:00 2001 From: Harudev0914 Date: Thu, 30 Jul 2026 14:04:20 +0900 Subject: [PATCH] docs: centre the README header block in all four languages Title, language switcher, tagline and badges were left-aligned and stacked. Now in

blocks, which is what the ecosystem expects at the top of a README and what makes the language switcher read as a switcher rather than as a line of prose. The tagline gains the two things a reader wants before deciding to keep reading -- what routes the models, and which surfaces exist -- and the badges gain ruff and mypy, both of which CI now enforces. Badge URLs move from relative to absolute. Relative paths survive a fork, but these are only meaningful pointing at this repository's Actions in the first place, and absolute renders correctly outside GitHub too. The licence badge is no longer a link: there is no LICENSE file to link to. pyproject declares Apache-2.0 but the licence text is not in the tree, which is worth fixing separately -- naming a copyright holder is not mine to guess. Also corrects a garbled clause in the Korean noise-estimation row. Co-Authored-By: Claude Opus 5 (1M context) --- README.ja.md | 33 +++++++++++++++++++++++---------- README.ko.md | 35 ++++++++++++++++++++++++----------- README.md | 34 +++++++++++++++++++++++----------- README.zh.md | 33 +++++++++++++++++++++++---------- 4 files changed, 93 insertions(+), 42 deletions(-) diff --git a/README.ja.md b/README.ja.md index a0bd633..b5b5e1a 100644 --- a/README.ja.md +++ b/README.ja.md @@ -1,13 +1,26 @@ -# VisionSR - -[English](README.md) · [한국어](README.ko.md) · [中文](README.zh.md) · **日本語** - -[![CI](../../actions/workflows/ci.yml/badge.svg)](../../actions/workflows/ci.yml) -[![Security](../../actions/workflows/security.yml/badge.svg)](../../actions/workflows/security.yml) -![Python](https://img.shields.io/badge/python-3.11%2B-blue) -![License](https://img.shields.io/badge/license-Apache--2.0-green) - -画像の拡大・復元・顔の再構成・ディテール復元を行う AI 超解像エンジン。 +

VisionSR

+ +

+ English · + 한국어 · + 中文 · + 日本語 +

+ +

+ AI 超解像エンジン · 画像の拡大・復元・顔の再構成・ディテール復元
+ モデル自動振り分け · 9 モデル · 4GB VRAM からのタイル推論 · 完全ローカル実行
+ CLI · HTTP API · Web UI · Electron デスクトップアプリ +

+ +

+ CI + Security + Python 3.11+ + Ruff + mypy + License: Apache-2.0 +

画像を渡すと、それが何なのか（写真・人物・線画・スキャン）、どの程度傷んでいるか （ノイズ・ブラー・JPEG）を自ら判断し、適した モデルを選び、実際に使える VRAM に合わせて diff --git a/README.ko.md b/README.ko.md index 74144b4..6e2b2a1 100644 --- a/README.ko.md +++ b/README.ko.md @@ -1,13 +1,26 @@ -# VisionSR - -[English](README.md) · **한국어** · [中文](README.zh.md) · [日本語](README.ja.md) - -[![CI](../../actions/workflows/ci.yml/badge.svg)](../../actions/workflows/ci.yml) -[![Security](../../actions/workflows/security.yml/badge.svg)](../../actions/workflows/security.yml) -![Python](https://img.shields.io/badge/python-3.11%2B-blue) -![License](https://img.shields.io/badge/license-Apache--2.0-green) - -이미지 업스케일링, 복원, 얼굴 복구, 디테일 재구성을 수행하는 AI 초해상도 엔진. +

VisionSR

+ +

+ English · + 한국어 · + 中文 · + 日本語 +

+ +

+ AI 초해상도 엔진 · 이미지 업스케일링, 복원, 얼굴 복구, 디테일 재구성
+ 모델 자동 라우팅 · 모델 9종 · 4GB VRAM부터 타일 추론 · 완전 로컬 실행
+ CLI · HTTP API · 웹 UI · Electron 데스크톱 앱 +

+ +

+ CI + Security + Python 3.11+ + Ruff + mypy + License: Apache-2.0 +

이미지를 넣으면 그것이 무엇인지(사진, 인물, 선화, 스캔), 얼마나 손상됐는지(노이즈, 블러, JPEG)를 스스로 판단해 알맞은 모델을 고르고, 실제 가용 VRAM에 맞춰 타일 크기를 @@ -138,7 +151,7 @@ tests/ 두 계층: 무표식(가중치·GPU 불필요)과 @pyte | 구성 요소 | 방법 | 왜 신경망이 아닌가 | |---|---|---| | 콘텐츠 분류 | 색상 수·채도·엣지·주파수 통계에 대한 휴리스틱 앙상블 | 설명 가능 — `ImageAnalysis.scores`가 어떤 신호가 결정했는지 알려줌. 가중치도 GPU도 불필요하고 약 10ms. | -| 노이즈 추정 | Immerkær의 Laplacian-of-Laplacian 커널 | 무참조 해석해라 학습 데이터가 필요 없음. | +| 노이즈 추정 | Immerkær의 Laplacian-of-Laplacian 커널 | 무참조이고 해석적으로 닫힌 형태라 학습 데이터가 필요 없음. | | 블러 추정 | Laplacian 분산 + 그래디언트 통계 | 위와 동일. | | JPEG 아티팩트 추정 | 8x8 격자 경계의 불연속 측정 | 추론이 아니라 아티팩트를 직접 측정. | | 모델 선택 | 위 지표에 대한 임계값 규칙 + 우선순위 | 잘못된 라우팅이 가장 눈에 띄는 실패이므로 설명 가능해야 함. | diff --git a/README.md b/README.md index b472856..0974d1f 100644 --- a/README.md +++ b/README.md @@ -1,14 +1,26 @@ -# VisionSR - -**English** · [한국어](README.ko.md) · [中文](README.zh.md) · [日本語](README.ja.md) - -[![CI](../../actions/workflows/ci.yml/badge.svg)](../../actions/workflows/ci.yml) -[![Security](../../actions/workflows/security.yml/badge.svg)](../../actions/workflows/security.yml) -![Python](https://img.shields.io/badge/python-3.11%2B-blue) -![License](https://img.shields.io/badge/license-Apache--2.0-green) - -Enterprise AI super resolution: image upscaling, restoration, face recovery and -intelligent detail reconstruction. +

VisionSR

+ +

+ English · + 한국어 · + 中文 · + 日本語 +

+ +

+ Enterprise AI super resolution · upscaling, restoration, face recovery and detail reconstruction
+ Automatic model routing · 9 models · tiled inference from 4 GB VRAM · fully local
+ CLI · HTTP API · Web UI · Electron desktop app +

+ +

+ CI + Security + Python 3.11+ + Ruff + mypy + License: Apache-2.0 +

Give it an image; it works out what the image *is* (photograph, portrait, line art, scan), how badly it is damaged (noise, blur, JPEG), picks the model that suits, sizes diff --git a/README.zh.md b/README.zh.md index 00bb65d..717cc73 100644 --- a/README.zh.md +++ b/README.zh.md @@ -1,13 +1,26 @@ -# VisionSR - -[English](README.md) · [한국어](README.ko.md) · **中文** · [日本語](README.ja.md) - -[![CI](../../actions/workflows/ci.yml/badge.svg)](../../actions/workflows/ci.yml) -[![Security](../../actions/workflows/security.yml/badge.svg)](../../actions/workflows/security.yml) -![Python](https://img.shields.io/badge/python-3.11%2B-blue) -![License](https://img.shields.io/badge/license-Apache--2.0-green) - -企业级 AI 超分辨率引擎：图像放大、修复、人脸还原与细节重建。 +

VisionSR

+ +

+ English · + 한국어 · + 中文 · + 日本語 +

+ +

+ 企业级 AI 超分辨率引擎 · 图像放大、修复、人脸还原与细节重建
+ 模型自动路由 · 9 个模型 · 4GB 显存起的分块推理 · 完全本地运行
+ 命令行 · HTTP API · Web 界面 · Electron 桌面应用 +

+ +

+ CI + Security + Python 3.11+ + Ruff + mypy + License: Apache-2.0 +

给它一张图，它会自行判断这是什么（照片、人像、线稿、扫描件）、损坏到什么程度（噪点、 模糊、JPEG 压缩），据此选择合适的模型，并按实际可用显存决定分块大小后执行。你可以手动