CheckCardBoxAndDefectServer/app/utils/score_inference/CardScorer.py

import json
from typing import List, Dict, Any, Union
from app.core.logger import get_logger

logger = get_logger(__name__)


class CardScorer:
    """
    它从一个JSON配置文件加载评分规则，并根据输入的卡片数据计算分数。
    """

    def __init__(self, config_path: str):
        try:
            with open(config_path, 'r', encoding='utf-8') as f:
                self.config = json.load(f)
            self.base_score = self.config.get("base_score", 10.0)
        except FileNotFoundError:
            raise ValueError(f"配置文件未找到: {config_path}")
        except json.JSONDecodeError:
            raise ValueError(f"配置文件格式错误: {config_path}")

    @staticmethod
    def _get_score_from_tiers(value: float, rules: List[Dict[str, Any]]) -> float:
        """
        根据新的区间规则查找值对应的分数。
        格式: [{"min": 0, "max": 0.5, "deduction": -1}, ...]
        逻辑: min <= value < max (最后一项如果是inf，则包含)
        """
        for tier in rules:
            # 获取范围，处理可能的字符串
            min_val = tier.get("min", 0)
            max_val_raw = tier.get("max", "inf")
            deduction = tier.get("deduction", 0.0)

            # 处理 inf
            if max_val_raw == "inf":
                max_val = float('inf')
            else:
                max_val = float(max_val_raw)

            min_val = float(min_val)

            # 判定范围：左闭右开 [min, max)
            if min_val <= value < max_val:
                return float(deduction)

        # 如果数值非常大（超过了所有定义的max），或者没有匹配到
        if rules:
            return float(rules[-1].get("deduction", 0))
        return 0.0

    def calculate_defect_score(self,
                               card_defect_type: str,
                               card_aspect: str,
                               defect_data: Dict,
                               is_write_score: bool = True) -> Union[float, dict]:
        """
        一个通用的缺陷计分函数，用于计算角、边、表面的加权分数。
        card_defect_type: 'corner', 'edge', 'face'
        card_aspect: 为 front或 back
        is_write_score: 是否将分数写入json并返回
        """
        if card_defect_type != "corner" and card_defect_type != "edge" and card_defect_type != "face":
            raise TypeError("calculate_centering_score:card_type 只能为 'corner', 'edge', 'face'")
        if card_aspect != "front" and card_aspect != "back":
            raise TypeError("calculate_defect_score:card_type 只能为 front 或 back")

        aspect_config = self.config[card_defect_type]
        total_deduction = 0.0
        weighted_scores = {}

        # 1. 计算每种缺陷类型的总扣分
        for defect in defect_data['defects']:
            if defect['defect_type'] != card_defect_type:
                continue

            if card_defect_type == 'corner' or card_defect_type == 'edge':
                if defect['label'] in ['wear', 'wear_and_impact', 'wear_and_stain']:
                    defect_type = "wear_area"
                elif defect['label'] in ['impact', 'damaged']:
                    defect_type = "loss_area"
                else:
                    logger.error(f"数据缺陷类型不存在: {defect['label']}")
                    raise TypeError(f"数据缺陷类型不存在: {defect['label']}")
            else:
                if defect['label'] in ['wear', 'wear_and_impact', 'damaged']:
                    defect_type = "wear_area"
                elif defect['label'] in ['scratch', 'scuff']:
                    defect_type = "scratch_length"
                elif defect['label'] in ['pit', 'impact', 'protrudent']:
                    defect_type = "pit_area"
                elif defect['label'] in ['stain']:
                    defect_type = "stain_area"
                else:
                    logger.error(f"数据缺陷类型不存在: {defect['label']}")
                    raise TypeError(f"数据缺陷类型不存在: {defect['label']}")

            # 获取规则列表
            rules = aspect_config['rules'].get(defect_type)
            if not rules:
                logger.error(f"计算分数过程, 未找到配置规则: {defect_type}")
                raise KeyError(f"计算分数过程, 未找到配置规则: {defect_type}")

            # 对于划痕取长度, 其他取面积
            if defect_type == "scratch_length":
                area_mm = max(defect['width'], defect['height'])
                defect['scratch_length'] = area_mm
            else:
                area_mm = defect['actual_area']
            # 累加所有同类型缺陷的扣分
            if defect_type not in weighted_scores.keys():
                weighted_scores[defect_type] = 0

            # 计算单个缺陷扣分
            the_score = self._get_score_from_tiers(area_mm, rules)
            print(f"[{card_defect_type}, {defect_type}]: {area_mm}, {the_score}")
            weighted_scores[defect_type] += the_score

            # 将分数写入json
            if is_write_score:
                if "score" not in defect:
                    # 新建的时候
                    logger.info(f"新建分数score: {the_score}")
                    defect['score'] = the_score
                    defect["new_score"] = None
                elif defect.get("new_score") is None:
                    # 初次修改
                    # if defect["score"] != the_score:
                    #     logger.info(f"初次修改 -> new_score: {the_score} (原score: {defect['score']})")
                    if defect.get("edit_type") == "edit" or defect.get("edit_type") == "add":
                        defect["new_score"] = the_score
                    elif defect.get("edit_type") == "del":
                        logger.info(f"del {defect_type} 补回: {the_score}")
                        defect["new_score"] = 0
                        weighted_scores[defect_type] -= the_score

                elif "score" in defect and defect["new_score"] is not None:
                    # 多次修改
                    # if defect["new_score"] != the_score:
                    #     defect["score"] = defect["new_score"]
                    if defect.get("edit_type") == "edit":
                        defect["new_score"] = the_score
                    elif defect.get("edit_type") == "del":
                        logger.info(f"del {defect_type} 补回: {the_score}")
                        defect["new_score"] = 0
                        weighted_scores[defect_type] -= the_score
                else:
                    defect['score'] = the_score
                    defect["new_score"] = None

        # 2. 根据权重/系数计算最终扣分
        weights = aspect_config.get(f"{card_aspect}_weights") or aspect_config.get("coefficients")
        if not weights:
            raise ValueError(f"在配置中未找到 '{card_defect_type}' 的权重/系数")

        print(weighted_scores)
        for defect_type, score in weighted_scores.items():
            total_deduction += score * weights.get(defect_type, 1.0)

        final_weights = aspect_config["final_weights"][card_aspect]
        final_score = total_deduction * final_weights
        logger.info(f"final weights: {final_weights}, final score: {final_score}")

        if is_write_score:
            defect_data[f"{card_aspect}_{card_defect_type}_deduct_score"] = final_score
            return defect_data
        else:
            return final_score

    def calculate_centering_score(self,
                                  card_aspect: str,
                                  center_data: dict,
                                  is_write_score: bool = False) -> Union[float, dict]:
        """
        计算居中度分数。
        card_type 为 front或 back
        is_write_score: 是否将分数写入json并返回
        """
        if card_aspect != "front" and card_aspect != "back":
            raise TypeError("calculate_centering_score:card_type 只能为 front 或 back")

        centering_config = self.config['centering'][card_aspect]
        rules = centering_config['rules']
        coefficients = centering_config['coefficients']

        center_left = center_data['box_result']['center_inference']['center_left']
        center_right = center_data['box_result']['center_inference']['center_right']
        center_top = center_data['box_result']['center_inference']['center_top']
        center_bottom = center_data['box_result']['center_inference']['center_bottom']

        # 将比例转换为用于查找规则的单个最大值
        h_lookup_val = max(center_left, center_right)
        v_lookup_val = max(center_top, center_bottom)

        h_deduction = self._get_score_from_tiers(h_lookup_val, rules) * coefficients['horizontal']
        v_deduction = self._get_score_from_tiers(v_lookup_val, rules) * coefficients['vertical']

        print(h_deduction, v_deduction)
        final_weight = self.config['centering']["final_weights"][card_aspect]
        final_score = (h_deduction + v_deduction) * final_weight

        logger.info(f"final weight: {final_weight}, final score: {final_score}")

        if is_write_score:
            center_data['deduct_score'] = final_score
            return center_data
        else:
            return final_score

    def formate_one_card_result(self, center_result: dict,
                                defect_result: dict,
                                card_defect_type: str,
                                card_aspect: str,
                                imageHeight: int,
                                imageWidth: int):
        try:
            # 获取计算总分的权重
            card_config = self.config['card']['PSA']

            # 计算各部分的最后分数
            # 计算居中
            final_center_score = None

            if card_defect_type == "corner_edge":
                center_score = center_result['deduct_score']
                center_weight = card_config['center']
                final_center_score = center_score * center_weight

                corner_score = defect_result[f"{card_aspect}_corner_deduct_score"]
                edge_score = defect_result[f"{card_aspect}_edge_deduct_score"]

                corner_weight = card_config['corner']
                edge_weight = card_config['edge']

                final_defect_score = corner_score * corner_weight + edge_score * edge_weight
                _used_compute_deduct_score = final_center_score + final_defect_score
                card_score = self.base_score + final_center_score + final_defect_score
            else:
                face_score = defect_result[f"{card_aspect}_face_deduct_score"]
                face_weight = card_config['face']

                final_defect_score = face_score * face_weight
                _used_compute_deduct_score = final_defect_score

                card_score = self.base_score + final_defect_score

        except Exception as e:
            logger.error(f"formate_one_card_result 从json获取分数失败: {e}")
            raise e

        data = {
            "result": {
                "center_result": center_result,
                "defect_result": defect_result,
                "imageHeight": imageHeight,
                "imageWidth": imageWidth,
                "card_center_deduct_score": final_center_score,
                "card_defect_deduct_score": final_defect_score,
                "_used_compute_deduct_score": _used_compute_deduct_score,
                "card_score": card_score
            }
        }
        return data


if __name__ == '__main__':
    # 1. 初始化评分器，加载规则
    scorer = CardScorer(r"C:\Code\ML\Project\CheckCardBoxAndDefectServer\app\core\scoring_config.json")
    # rulers = scorer.config['corners']['rules']['wear_area']
    #
    # score = scorer._get_score_from_tiers(0.06, rulers)
    # print(score)
    # print()

    # 居中分数
    center_data_path = r"C:\Code\ML\Project\CheckCardBoxAndDefectServer\_temp_work\pokemon_card_center-center_result.json"
    with open(center_data_path, 'r', encoding='utf-8') as f:
        center_data = json.load(f)
    center_data = scorer.calculate_centering_score("front", center_data, True)
    print(center_data)

    # 边角分数
    edge_corner_data_path = r"C:\Code\ML\Project\CheckCardBoxAndDefectServer\_temp_work\pokemon_front_corner_no_reflect_defect-corner_result.json"
    with open(edge_corner_data_path, 'r', encoding='utf-8') as f:
        edge_corner_data = json.load(f)

    corner_data = scorer.calculate_defect_score("corner", 'front', edge_corner_data, True)
    print(corner_data)

    score = scorer.calculate_defect_score("edge", 'front', edge_corner_data, True)
    print(score)

    # 面分数
    face_data_path = r"C:\Code\ML\Project\CheckCardBoxAndDefectServer\_temp_work\pokemon_front_face_no_reflect_defect-face_result.json"
    with open(face_data_path, 'r', encoding='utf-8') as f:
        face_data = json.load(f)

    score = scorer.calculate_defect_score("face", 'front', face_data, True)
    print(score)