import type { InferenceSession } from "onnxruntime-common";
import type { CoreCanvas, PlatformProvider } from "../core/platform.js";
import { BaseAnalysis } from "./base.interface.js";
/**
 * Configuration options for face spoofing detection models
 */
export interface SpoofingOptions {
    /** Threshold score values for spoofing detection */
    threshold: number;
    /** Wether to activate the anti-spoofing analysis */
    enable: boolean;
}
/**
 * Spoofing detection result
 */
export interface SpoofingResult {
    /** Whether the image is real or fake */
    real: boolean;
    /** The given score from analysis (confidence of the predicted label) */
    score: number;
    /** Probability of the image being real (0-1) */
    realness: number;
    /** Probability of the image being fake (0-1) */
    fakeness: number;
}
export declare class SpoofingDetection extends BaseAnalysis {
    protected className: string;
    protected firstModelPath: string;
    protected secondModelPath: string;
    protected firstSession: InferenceSession | null;
    protected secondSession: InferenceSession | null;
    protected options: SpoofingOptions;
    constructor(options?: Partial<SpoofingOptions>, platform?: PlatformProvider);
    initialize(): Promise<void>;
    /**
     * Checks if the model has been initialized
     * @returns True if initialized
     */
    protected isInitialized(): boolean;
    /**
     * Analyze whether a face is spoofed
     * @param image The image (full image recommended)
     * @param facialArea Optional facial bounding box {x, y, width, height}.
     *                   If not provided, uses the entire image as the facial area.
     */
    analyze(image: ArrayBuffer | CoreCanvas, facialArea?: {
        x: number;
        y: number;
        width: number;
        height: number;
    }): Promise<SpoofingResult>;
    /**
     * Preprocess image by cropping from original image with scale and resizing
     */
    protected preprocessPromise(canvas: CoreCanvas, bbox: {
        x: number;
        y: number;
        width: number;
        height: number;
    }, scale: number, outW: number, outH: number, session: InferenceSession): Promise<InferenceSession.OnnxValueMapType>;
    /**
     * Preprocess image by cropping from original image with scale and resizing
     */
    protected preprocessWithCrop(canvas: CoreCanvas, bbox: {
        x: number;
        y: number;
        width: number;
        height: number;
    }, scale: number, outW: number, outH: number): Float32Array;
    /**
     * Simple preprocess by resizing only (for pre-cropped faces)
     */
    protected preprocess(canvas: CoreCanvas, outW: number, outH: number): Float32Array;
    /**
     * Convert canvas to tensor in CHW format
     */
    protected canvasToTensor(canvas: CoreCanvas, width: number, height: number): Float32Array;
    /**
     * Calculate new bounding box with scale
     */
    protected getNewBox(srcW: number, srcH: number, bbox: {
        x: number;
        y: number;
        width: number;
        height: number;
    }, scale: number): {
        x: number;
        y: number;
        width: number;
        height: number;
    };
    /**
     * Run inference on a model
     */
    protected inference(session: InferenceSession, tensor: Float32Array, shape: number[]): Promise<InferenceSession.OnnxValueMapType>;
    /**
     * Apply softmax to logits
     */
    protected softmax(logits: Float32Array): Float32Array;
    /**
     * Postprocess model outputs
     */
    protected postprocess(firstOutput: InferenceSession.OnnxValueMapType, secondOutput: InferenceSession.OnnxValueMapType): SpoofingResult;
    destroy(): Promise<void>;
}