Skip to content

Latest commit

 

History

15 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

VisionText: High-Performance Offline OCR Framework for Android

VisionText is an advanced Optical Character Recognition (OCR) solution engineered for secure, real-time text extraction on edge devices. By leveraging a hybrid integration of Google ML Kit and optimized recognition engines, the system achieves sub-second latency for text detection and extraction without requiring an active network connection.


Technical Architecture

The framework is built to address the critical requirement for privacy-compliant data processing. VisionText ensures that 100% of image analysis occurs locally on the hardware, eliminating data transit risks.

Core Capabilities

  • On-Device Intelligence: Zero-latency text extraction using local machine learning models.
  • Real-Time Spatial Mapping: Dynamic coordinate-based mapping of text blocks with immediate visual feedback via high-precision overlays.
  • Granular Data Selection: Interactive selection logic allowing users to isolate and extract specific data points.
  • Resource Efficiency: Optimized for minimal CPU and memory overhead.

Performance & Output Analysis

The following data represents the system's accuracy in spatial mapping and hardware resource management.

System Visualization & Extraction

Detection Phase Extraction Result Selection UI
Mapping Extraction UI

Advanced Processing

Multi-Line Handling Document Segmentation Logic Verification
Lines Segments Logic

Resource Management

The system is optimized for edge computing. Below is the recorded CPU utilization during a high-density text scan: CPU Utilization


Implementation Pipeline

The application manages the OCR lifecycle through a specialized six-stage technical pipeline:

  1. User Interface Layer: Streamlined interface implemented via activity_main.xml.
  2. Asynchronous Image Acquisition: Integration of the CameraX API for high-resolution capture.
  3. Neural Text Recognition: Utilization of the Google ML Kit Text Recognizer.
  4. Spatial Visualization: Custom TextOverlayView.kt for real-time rendering.
  5. Interactive Event Handling: Event-driven touch listeners for text selection.
  6. Data Output Management: Automated string processing for clipboard export.

Technical Specifications

  • Language: Kotlin 1.9+
  • Minimum SDK: API Level 24 (Android 7.0)
  • Target SDK: API Level 34
  • Core Dependencies: Google ML Kit, CameraX, Kotlin Coroutines

Project Origins

This system was developed as a collaborative initiative by a team of six developers focused on edge-computing and mobile machine learning implementations.

About

A high-performance, privacy-focused Android OCR library for real-time text extraction. Works 100% offline using Google ML Kit and Tesseract. Perfect for scanning documents, receipts, and notes without data leaving the device. Fast, lightweight, and easy to integrate.

Topics

Resources

Stars

2 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages