import * as poseDetection from '@tensorflow-models/pose-detection';
import { IAgentRuntime } from '@elizaos/core';
import { DetectedObject, PersonInfo } from './types';
export interface VisionModelConfig {
    enableObjectDetection?: boolean;
    enablePoseDetection?: boolean;
    florence2?: {
        baseUrl: string;
        apiKey?: string;
    };
    vlm?: {
        model: string;
        temperature: number;
        maxTokens: number;
    };
}
export type Pose = 'sitting' | 'standing' | 'lying' | 'walking' | 'unknown';
export interface PoseLandmark {
    name: string;
    x: number;
    y: number;
    score: number;
}
export declare class VisionModels {
    private runtime;
    private config;
    private objectDetectionModel;
    private poseDetector;
    private initialized;
    private florence2Model;
    private cocoSsdModel;
    private posenetModel;
    constructor(runtime: IAgentRuntime, config?: VisionModelConfig);
    initialize(config: VisionModelConfig): Promise<void>;
    hasObjectDetection(): boolean;
    hasPoseDetection(): boolean;
    detectObjects(imageData: Buffer, width: number, height: number, description?: string): Promise<DetectedObject[]>;
    private enhancedObjectDetection;
    private generatePlausibleBoundingBox;
    detectPoses(imageData: Buffer, width: number, height: number, description?: string): Promise<PersonInfo[]>;
    private enhancedPoseDetection;
    private generatePlausibleKeypoints;
    convertPosesToPersonInfo(poses: poseDetection.Pose[]): PersonInfo[];
    private determinePoseFromKeypoints;
    private determineFacingDirection;
    dispose(): Promise<void>;
}