ppu-paddle-ocr/core/base-detection.service.d.ts

Lightweight, probably the fastest PaddleOCR SDK in TypeScript. Runs anywhere JavaScript runs: Node.js, Bun, Deno, mobile react-native, web browsers, and browser extensions. Docker & CLI supported. The official SDK is browser-only. Accurate text detection

import type { InferenceSession } from "onnxruntime-common";
import type { Box, DebuggingOptions, DetectionOptions, ProcessingEngine } from "../interface.js";
import type { CoreCanvas, PlatformProvider } from "./platform.js";
/**
 * Result of preprocessing an image for text detection.
 *
 * Contains the normalized float tensor, dimensions, and scale factors
 * needed to map detection output back to original image coordinates.
 */
export type PreprocessDetectionResult = {
    /** Normalized float tensor (CHW layout, 3 channels). */
    tensor: Float32Array;
    /** Width of the padded/resized tensor in pixels. */
    width: number;
    /** Height of the padded/resized tensor in pixels. */
    height: number;
    /** Scale factor applied during resize (`resized / original`). */
    resizeRatio: number;
    /** Original image width before preprocessing. */
    originalWidth: number;
    /** Original image height before preprocessing. */
    originalHeight: number;
};
/**
 * Service for detecting text regions in images
 */
export declare class BaseDetectionService {
    protected readonly options: DetectionOptions;
    protected readonly debugging: DebuggingOptions;
    protected readonly session: InferenceSession;
    protected readonly platform: PlatformProvider;
    protected readonly engine: ProcessingEngine;
    private lastDetectionCanvas;
    constructor(platform: PlatformProvider, session: InferenceSession, options?: Partial<DetectionOptions>, debugging?: Partial<DebuggingOptions>, engine?: ProcessingEngine);
    /**
     * Logs a message if verbose debugging is enabled
     */
    protected log(message: string): void;
    /**
     * Main method to run text detection on an image
     * @param image ArrayBuffer of the image or platform-specific Canvas
     */
    run(image: ArrayBuffer | CoreCanvas): Promise<Box[]>;
    /**
     * Preprocess an image for text detection
     */
    private preprocessDetection;
    /**
     * Run the detection model inference
     */
    private runInference;
    /**
     * Process detection results to extract bounding boxes
     */
    private postprocessDetection;
    /**
     * Post-process detection using OpenCV contours (v4-compatible, more accurate)
     */
    private postprocessWithOpenCV;
    /**
     * Post-process detection using canvas-native region detection
     */
    private postprocessWithCanvasNative;
    /**
     * Debug the detection canvas in binary image format (thresholded)
     */
    private debugDetectionCanvas;
    /**
     * Debug the bounding boxes by drawing a rectangle onto the original image
     */
    private debugDetectedBoxes;
}