yolov8缺陷检测改进步骤

yolov8改进步骤

1.看视频:parse
2.修改fitness()函数
位置：ultralytics/utils/metrics.py 检索fitness(self)

def fitness(self):
    """Model fitness as a weighted combination of metrics."""
    w = [0.0, 1.0, 0.0, 0.0]  # weights for [P, R, mAP@0.5, mAP@0.5:0.95]
    return (np.array(self.mean_results()) * w).sum()
1
2
3
4

作用：修改报错best.py的依据是百分百看recall召回率
3.创建dataset/data.yaml文件
把path改成data的绝对路径地址！！！

path: /public/home/test202306/zj/data
train: train
val: val
test:

# Classes
names:
  0: debris on the front of the vehicle
  1: cover open
  2: layer detachment
  3: anti loosening wire breakage
  4: oil leakage
1
2
3
4
5
6
7
8
9
10
11
12

4.创建train.py

import warnings
warnings.filterwarnings('ignore')
from ultralytics import YOLO

if __name__ == '__main__':
    str = 'yolov8-C2f-DCNv3.yaml'
    model = YOLO('ultralytics/cfg/models/Add/{}'.format(str))
    # model.load('yolov8n.pt') # 是否加载预训练权重,科研不建议大家加载否则很难提升精度
    model.train(data=r'dataset/data.yaml',
                cache=False,
                imgsz=640,
                epochs=200,
                single_cls=False,  # 是否是单类别检测
                batch=128,
                close_mosaic=10,
                workers=6,
                device='0',
                optimizer='SGD', # using SGD
                # resume='runs/train/exp21/weights/last.pt', # 如过想续训就设置last.pt的地址
                amp=False,  # 如果出现训练损失为Nan可以关闭amp
                project='runs/train',
                name='exp_{}'.format(str[:-5]),  # 当前实验的名称
                )
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23