Skip to content

README_EN.md

Latest commit

 

History

History
146 lines (108 loc) · 6.9 KB

README_EN.md

File metadata and controls

146 lines (108 loc) · 6.9 KB

PaddleOCR4Android

Offline OCR App for Android — Powered by PaddleOCR

A user-friendly Android application wrapping the powerful PaddleOCR engine. 100% offline — supports real-time text recognition in Chinese, English, Japanese, Korean, and Traditional Chinese.

PaddleOCR4Android Screenshot

中文文档

✨ Highlights

  • Fully Offline: All recognition runs locally — no internet needed, privacy-friendly, works anywhere.
  • PP-OCRv5 Engine: Latest-gen OCR model with high accuracy for Chinese, English, Japanese, and Traditional Chinese in a single model.
  • Smart Interaction: Three modes — Box Selection, Line-by-Line, and Word Segmentation — covering everything from paragraph extraction to precise word picking.
  • Ready to Use: APK bundles all language models (~61MB). Install and go.
  • Developer Friendly: Clean modular architecture and source code for easy integration into your own projects.

📥 Download

Download the latest APK and install it on your Android device.

Version Download Highlights
v2.7.0 (Latest) PaddleOCR-v2.7.0.apk Word segmentation upgrade: drag-to-select & auto-scroll
v2.6.3 PaddleOCR-v2.6.3.apk Optimized long-text word order & box selection copy order
v2.6.1 PaddleOCR-v2.6.1.apk Box selection drag & flow layout for word segmentation
v2.3.0 PaddleOCR-v2.3.0.apk Core engine upgraded to PP-OCRv5

Minimum requirement: Android 7.0 (API 24)

🌍 Supported Languages

Language Group Model Notes
Chinese / English / Japanese / Traditional PP-OCRv5 Default — single model, high accuracy
Korean (한국어) PP-OCRv3 ~60% accuracy, PP-OCRv5 integration pending

🛠️ Features

Core Recognition

  1. Offline Recognition: No network dependency — works anytime, anywhere.
  2. Multi-language Switching: One-tap language switching in settings, auto-loads the corresponding model.
  3. Smart Chinese Word Segmentation: Integrated Jieba segmentation engine with semantic-aware tokenization that preserves reading order.

Three Interaction Modes

Mode How to Use Best For
🖼️ Box Selection Drag a rectangle on the image to select all text within the area. Quickly grab irregularly laid-out text blocks.
📝 Line-by-Line Tap any text line to select it. Multi-tap for cumulative selection. Fast copy of multi-line paragraphs.
🔗 Word Segmentation Results shown as flow-style word chips. Tap to select, or horizontal drag for continuous selection. Precise word/phrase copy, similar to mobile WPS.

Selected text is highlighted in blue on the image in real-time, with results shown in the bottom panel.

Utilities

  • Copy & Share: One-tap copy to clipboard or share to other apps.
  • History: Auto-saves recognition history. View, delete, or clear anytime.
  • Instant Reset: Auto-clears content on back press, ready for next scan.

⚙️ Tech Stack

Component Technology Purpose
OCR Engine PaddleOCR + Paddle Lite Core recognition & mobile inference
Android Wrapper paddleocr4android Android API for PaddleOCR
Chinese Segmentation houbb/segment (Jieba lexicon) Smart Chinese word segmentation
Data Persistence Room Recognition history storage
UI Material Design 3 + AndroidX Modern, beautiful UI components

📂 Build from Source

# 1. Clone the repository
git clone https://github.com/jangviktor-web/PaddleOCR4Android.git
cd PaddleOCR4Android

# 2. Open in Android Studio or build with Gradle
./gradlew assembleRelease

# 3. Output APK:
# app/build/outputs/apk/release/app-release.apk

🏗️ Project Structure

├── app/                          # Main app module
│   └── src/main/
│       ├── java/.../             # Kotlin source
│       │   ├── MainActivity.kt     # Main UI & core logic
│       │   ├── MoreActivity.kt     # History & about page
│       │   ├── OcrOverlayView.kt   # Image touch interaction layer
│       │   └── FlowLayout.kt       # Word segmentation flow layout
│       ├── assets/models/          # Pre-bundled OCR models (.nb)
│       └── res/                    # Layouts, icons, themes
├── PaddleOCR4Android/              # PaddleOCR SDK module
└── build.gradle                    # Global build config

❤️ Credits

Built on these excellent open-source projects:

📜 Changelog

v2.7.0 (2026-05-17) — Word Segmentation Revolution
  • Drag-to-select words: Horizontal drag in the word segmentation panel to continuously select word chips.
  • FlowLayout auto-scroll: Panel auto-scrolls when dragging beyond boundaries.
  • Fixed: Long-text word segmentation losing selection state on scroll.
  • Fixed: Gesture conflict between drag selection and page scrolling.
v2.6.3 (2026-05-17) — Order Fix
  • Fixed: Word order scrambled in segmentation mode for long texts (e.g., classical Chinese essays). Now uses per-line independent segmentation.
  • Fixed: Box selection copy order now follows top-to-bottom, left-to-right reading order.
v2.6.1 (2026-05-17) — Interaction Redesign
  • Box selection drag: Drag to draw a rectangle, batch-select all text in the area.
  • Flow segmentation: Word layout upgraded to WPS-style flow chips — more intuitive and beautiful.
  • Fixed: Copied text appearing with one character per line.
v2.3.0 (2026-05-16) — Engine Upgrade
  • Upgraded OCR engine from PP-OCRv4 to PP-OCRv5 — smaller model, higher accuracy, wider language coverage.
  • APK size reduced from 102MB to 61MB.

📄 License

This project is licensed under the MIT License.