yolo8训练自己的模型

王大锤4391

已于 2024-09-18 11:04:14 修改

阅读量239

点赞数 1

文章标签： YOLO

于 2024-09-13 15:13:57 首次发布

本文链接：https://blog.csdn.net/weixin_41037490/article/details/142211724

版权

1.数据源准备

1.1 准备图片资源

1.2 对图片资源标注，生成对应的 .txt 文件，里面的数字表示物体被标注的 x或y 等坐标点信息

1.2.1 标注工具下载以及使用教程参考

Windows 10下安装labelImg标注工具！_labelimg windows exe 1.5版本-CSDN博客

1.2.3 标注工具使用教程参考

如何使用labelImg标注数据集，最详细的深度学习标签教程_labelimg使用教程-CSDN博客

2.标注 yolo 的文件

key.txt

0 0.460907 0.492033 0.848319 0.605170
0 0.451525 0.485305 0.807662 0.578966
0 0.462080 0.493980 0.800625 0.588527
0 0.453870 0.488314 0.834246 0.593484

3.标签配置文件

train: D:/pyourch-workspace/yoloy8-project/datasets/coco8/images/train
val: D:/pyourch-workspace/yoloy8-project/datasets/coco8/images/val

nc: 5
names: ['cigarette_case',  'key', 'pen', 'smoky', 'watch']
#        烟盒              钥匙     笔      香烟      手表

3.模型训练主要代码

from ultralytics import YOLO

##TODO yolov8 模型训练

# Load a COCO-pretrained YOLOv8n model
## 加载预训练模型
#第三种:根据yaml文件构建一个新模型，然后将预训练模型的参数转移到新模型中，然后进行训练，对YOLO8网络进行改进的适合选用此种训练方式，而且训练时间不至于过长
model = YOLO("yolov8n.yaml").load('yolov8n.pt')  # build from YAML and transfer weights

# Train the model on the COCO8 example dataset for 100 epochs
#epochs指定训练多少轮
#imgsz指定图片大小
#results = model.train(data="D:/pyourch-workspace/yoloy8-project/datasets/coco8/data.yaml", epochs=100, imgsz=640)
results = model.train(data="D:/pyourch-workspace/yoloy8-project/datasets/coco8/data.yaml",imgsz=640, epochs=100)

4. 训练结果集

5.使用自己训练的模型进行推理

from ultralytics import YOLO
import cv2
model = YOLO("D:/pyourch-workspace/yoloy8-project/runs/detect/train17/weights/best.pt")

results  = model(['D:/pyourch-workspace/yoloy8-project/image/100.jpg'])

###使用自己训练的模型  对   图片进行推理
#model.predict('D:/pyourch-workspace/yoloy8-project/image/group.jpg', save=True, imgsz=640, conf=0.5)


# Process results list
for result in results:
    boxes = result.boxes  # Boxes object for bounding box outputs
    masks = result.masks  # Masks object for segmentation masks outputs
    keypoints = result.keypoints  # Keypoints object for pose outputs
    probs = result.probs  # Probs object for classification outputs
    obb = result.obb  # Oriented boxes object for OBB outputs
    result.show()  # display to screen
    result.save(filename="result-2024-09-14-12-02.jpg")  # save to disk