yolov8训练笔记

一、 Yolov8直接安装和conda安装
1.1安装
yolov8有两种安装方式，一种可直接安装U神的库，为了方便管理，还是在conda里面安装：
conda create -n yolov8 python=3.8
conda activate ylolv8
pip install ultralytics
1.2验证
安装完成之后，验证是否安装成功。
yolo task=segment mode=predict model=yolov8s-seg.pt source=‘1.jpg’ show=True

yolo predict model=yolov8n.pt source=‘https://ultralytics.com/images/bus.jpg’

YOLOV8目标识别——详细记录从环境配置、自定义数据、模型训练到模型推理部署 (xdnf.cn)

二、 Yolov8目标检测detect
2.1 yolov8s.pt预训练模型开始训练
数据集制作时，将数据存放在datasets里，从预训练的*.pt模型开始训练
yolo task=detect mode=train model=datasets/yolov8s.pt epochs=10 batch=-1 data=datasets/first/cow_data.yaml

yolo task=detect mode=train model=yolov8m.pt epochs=5000 batch=8 imgsz=640 data=A_dataset_fzz/cow_data.yaml

yolo detect mode=train model=yolov8m.pt epochs=3500 batch=8 imgsz=1280 data=A_dataset_fzz/cow_data.yaml

yolo task=segment mode=train model=yolov8m.pt epochs=3500 batch=16 imgsz=1280 data=A_dataset_fzz/cow_data.yaml

yolo task=detect mode=val split=test model=runs/detect/train3/weights/best.pt data=datasets/first/cow_data.yaml
2.2 coco.yaml模型直接训练
yolo task=detect mode=train model= /data/coding/fzz/ultralytics/ultralytics/cfg/models/v8/yolov8.yaml epochs=3500 batch=16 imgsz=1280 data=datasets/split_dir/myseg.yaml

2.3 模型预测
yolo predict model=yolov8n.pt imgsz=320 source=‘https://ultralytics.com/images/bus.jpg’
yolo predict model=yolov8n.pt imgsz=640 source=‘https://ultralytics.com/images/bus.jpg’
yolo predict model=yolov8n.pt imgsz=1280 source=‘https://ultralytics.com/images/bus.jpg’
适合于不同尺寸，如5120*5120

2.4中断继续训练
yolo train resume model=runs/detect/train3/weights/last.pt

2.5结果推理
from ultralytics import YOLO

model = YOLO(“runs/train/weights/best.pt”) # load a pretrained model (recommended for training)

path = model.export(format=“onnx”) # export the model to ONNX format

单张图片：
yolo detect predict model=runs/train/weights/best.onnx source=‘test_img/1.jpg’
增强后推理：
yolo detect predict model=runs/train3/weights/best.pt source=‘datasets/1.bmp’ conf=0.5 iou=0.5 augment
多张图片进行推理：
yolo detect predict model=runs/train/weights/best.onnx source=‘datasets/first/images/val/’