Swift package · v0.1.0

KLOCRPipeline

Run OCR stages with explicit geometry and model boundaries.

A staged OCR pipeline for image preprocessing, Vision text recognition, reading-order reconstruction, and optional PaddleOCR Core ML inference.

API documentation View source

Platforms: iOS 17+ · macOS 14+
Toolchain: Swift 6.0+
License: MIT
Dependencies: None

Functions

Explicit preprocessing and recognition stages
Vision and optional PaddleOCR paths
Reading-order reconstruction through KLReadingOrder
Testable model and runtime boundaries

Not included

It runs OCR stages and returns structured observations. Image acquisition, user-facing editing, model distribution, and persistence remain with the integrating application.

Install with SwiftPM

Package.swiftSwift

.package(url: "https://github.com/KoenLee1023/KLOCRPipeline.git", from: "0.1.0")

Usage

Integration exampleSwift

let result = try await pipeline.recognize(image)
let ordered = readingOrder.reconstruct(candidates: result.candidates)

API documentation

Open the complete API documentation.

KLOCRPipeline

Runs configured OCR stages and returns recognized candidates with source geometry.

KLOCRVision

Provides the Apple Vision recognition stage for supported image inputs.

PPOCRInferenceSession

Owns the boundary between validated tensors and a Core ML OCR model.

Demo apps

Pipeline Playground

Run staged recognition and inspect candidates and reading order.

Vision Text Lab

Compare Vision output with post-processing and layout reconstruction.