YOLOv8使用COCO评价指标，解决AssertionError: Results do not correspond to current coco set.

YOLO评测指标和COCO评测指标还是有些区别的，通过数据表明YOLO评测指标要比COCO评测指标高2,3个点都是正常的。
跟着流程走吧！！

yolo设置

在yolo.eval()评估的时候，需要设置save_json = True保存结果json文件

model = YOLO("weight/best.pt")
model.val(data="data.yaml",imgsz=640, save_json=True)

结果默认保存在runs/detect/val/predictions.json

Id转化

coco格式id是从0-n，一个序列。而yolo的id是文件名称，字符。需要借助标签文件进行映射，写到新文件中

def cover_pred_json_id(anno_json_path, pred_json_path):
    with open(anno_json_path, "r") as f:
        ann_json = json.load(f)
    with open(pred_json_path, "r") as f:
        pred_json = json.load(f)
    for pred_item in pred_json:
        img_id = pred_item["image_id"]
        ann_id = [ann_item["id"] for ann_item in ann_json["images"] if ann_item["file_name"][:-4] == img_id]
        try:
            pred_item["image_id"] = ann_id[0]
        except IndexError:
            print(img_id)

    out_json_path = os.path.join(os.path.dirname(pred_json_path),"newpred.json")
    with open(out_json_path, 'w') as file:
        json.dump(pred_json, file, indent=4)
    return out_json_path

评测

完整代码

def parse_opt():
    parser = argparse.ArgumentParser()
    parser.add_argument('--anno_json', type=str, default='datasets/annotations/instances_val2017.json', help='training model path')
    parser.add_argument('--pred_json', type=str, default='utils/json_files/newcocopred.json', help='data yaml path')
    
    return parser.parse_known_args()[0]

def cover_pred_json_id(anno_json_path, pred_json_path):
    with open(anno_json_path, "r") as f:
        ann_json = json.load(f)
    with open(pred_json_path, "r") as f:
        pred_json = json.load(f)
    for pred_item in pred_json:
        img_id = pred_item["image_id"]
        ann_id = [ann_item["id"] for ann_item in ann_json["images"] if ann_item["file_name"][:-4] == img_id]
        try:
            pred_item["image_id"] = ann_id[0]
        except IndexError:
            print(img_id)

    out_json_path = os.path.join(os.path.dirname(pred_json_path),"newpred.json")
    with open(out_json_path, 'w') as file:
        json.dump(pred_json, file, indent=4)
    return out_json_path

if __name__ == '__main__':
    opt = parse_opt()
    anno_json = opt.anno_json
    pred_json = opt.pred_json
    pred_json = cover_pred_json_bbox(anno_json, pred_json) # cover yolo id to coco id
    anno = COCO(anno_json)  # init annotations api
    print(pred_json)
    pred = anno.loadRes(pred_json)  # init predictions api
    eval = COCOeval(anno, pred, 'bbox')
    eval.evaluate()
    eval.accumulate()
    eval.summarize()

如果不出意外的话，会正确打印结果：

错误

1. 如果出现评测指标为0的时候，说明category_id都没对应上，coco是从1开始计算, 而yolo是0开始
   更改ultralytics/models/yolo/detect/val.py设置self.is_coco = True

    def init_metrics(self, model):
        """Initialize evaluation metrics for YOLO."""
        val = self.data.get(self.args.split, "")  # validation path
        self.is_coco = True # isinstance(val, str) and "coco" in val and val.endswith(f"{os.sep}val2017.txt")  # is COCO
        self.class_map = converter.coco80_to_coco91_class() if self.is_coco else list(range(1000))
        self.args.save_json |= self.is_coco  # run on final val if training COCO
        self.names = model.names
        self.nc = len(model.names)
        self.metrics.names = self.names
        self.metrics.plot = self.args.plots
        self.confusion_matrix = ConfusionMatrix(nc=self.nc, conf=self.args.conf)
        self.seen = 0
        self.jdict = []
        self.stats = dict(tp=[], conf=[], pred_cls=[], target_cls=[])

2. 如果出现AssertionError: Results do not correspond to current coco set的错误通常是预测文件和标签中的id值不对应，这时候就要查看预测的文件中是不是少了或者多了哪个文件。