如何计算整个数据集的mAP (平均平均精度)？

import numpy as np

def compute_ap(recall, precision):
    """计算AP值"""
    # 插值法计算PR曲线下的面积
    mrec = np.concatenate(([0.], recall, [1.]))
    mpre = np.concatenate(([0.], precision, [0.]))
    for i in range(mpre.size - 1, 0, -1):
        mpre[i - 1] = np.maximum(mpre[i - 1], mpre[i])
    i = np.where(mrec[1:] != mrec[:-1])[0]
    ap = np.sum((mrec[i + 1] - mrec[i]) * mpre[i + 1])
    return ap

def compute_map(ground_truth, predictions):
    """计算整个数据集的mAP"""
    classes = set([gt['class'] for gt in ground_truth])
    aps = []
    
    for cls in classes:
        cls_gt = [gt for gt in ground_truth if gt['class'] == cls]
        cls_pred = [pred for pred in predictions if pred['class'] == cls]
        
        # 根据预测得分排序
        cls_pred.sort(key=lambda x: x['score'], reverse=True)
        
        # 计算每个检测框的TP/FP，并生成recall和precision列表
        tp = [0] * len(cls_pred)
        fp = [0] * len(cls_pred)
        for i, pred in enumerate(cls_pred):
            # 这里需要实现具体的匹配逻辑（如IoU阈值）
            # ...
            tp[i] = ...  # 根据匹配结果设置TP/FP
            fp[i] = ...
        
        # 计算precision和recall
        precision = [sum(tp[:i+1]) / (sum(tp[:i+1]) + sum(fp[:i+1])) for i in range(len(tp))]
        recall = [sum(tp[:i+1]) / len(cls_gt) for i in range(len(tp))]
        
        ap = compute_ap(recall, precision)
        aps.append(ap)
    
    map = np.mean(aps)
    return map