cotton_color/YOLO
2024-11-13 18:06:33 +08:00
..
README.assets 更新README 2024-11-13 18:06:33 +08:00
yolov5-master Create 11.12.onnx 2024-11-12 15:14:22 +08:00
README.md 更新README 2024-11-13 18:06:33 +08:00

YOLO调试记录

data 数据yaml文件配置

path: ../datasets/dimo4
train: D:\yolov5-master\datasets\dimo4\images\train
val: D:\yolov5-master\datasets\dimo4\images\val
test: # test images (optional)
# Classes
names:
  0: dimo

模型导出

命令行形式

python export.py --weights runs/train/exp4/weights/best.pt --img 640 --batch 1 --device 0 --include onnx --opset 15       

weights地址: runs/train/exp4/weights/xxx.pt 路径中的模型为pt文件

在detect.py中batch为1 batch在detect中不能更改

image-20241113133535978

1

PyCharm配置

export.py
--weights
runs/train/exp4/weights/best.pt
--img
640
--device0
--include
onnx
--opset
15

模型加载

--weights
runs\train\exp4\weights\best.onnx
--data
D:\yolov5-master\data\dimo4.yaml

在detect.py中

输入的模型为在train.py中训练好的模型 :地址: runs/train/exp4/weights/xxx.pt

或者 runs/train/exp4/weights/xxx.onnx

image-20241113133825646

image-20241113133840896

image-20241113133850391