Yolov10训练自己的数据集

前置准备步骤

1. 代码下载
2. 数据集准备
   ----- datasets
   ----------- data.yml
   ----------- train
   ---------------------- images
   -------------------------------------------- 1.png
   ---------------------- labels
   -------------------------------------------- 1.txt
   ----------- val
   ---------------------- images
   -------------------------------------------- 1.png
   ---------------------- labels
   -------------------------------------------- 1.txt
3. txt文件内容
   label, center_x, center_y, w, h
   0, 0.5, 0.5, 0.3, 0.4

训练以及测试代码

训练代码

#coding:utf-8
from ultralytics import YOLOv10
# 模型配置文件
model_yaml_path = "ultralytics/cfg/models/v10/yolov10n.yaml"
#数据集配置文件
data_yaml_path = 'datasets/Data/data.yaml'
#预训练模型
pre_model_name = 'yolov10n.pt'

if __name__ == '__main__':
    #加载预训练模型
    model = YOLOv10(model_yaml_path).load(pre_model_name)
    #训练模型
    results = model.train(data=data_yaml_path,
                          epochs=20,
                          batch=4,
                          name='train_v10')

测试代码

from ultralytics import YOLOv10

# Load a pretrained YOLOv10n model
model = YOLOv10("yolov10n.pt")

# Perform object detection on an image
# results = model("test1.jpg")
results = model.predict("test1.jpg")

# Display the results
results[0].show()

c++ 部署

代码下载

onnx 库安装

https://github.com/microsoft/onnxruntime/releases/tag/v1.17.1 下载对应版本
cmake.txt配置相应路径

find_path(ONNXRUNTIME_INCLUDE_DIR onnxruntime_c_api.h
    HINTS /home/xxx/data/lib/onnx-1.17.1/include
)
find_library(ONNXRUNTIME_LIBRARY onnxruntime
    HINTS /home/xxx/data/lib/onnx-1.17.1/lib
)

模型转换

from ultralytics import YOLOv10

model_name = "xxx.pt"
onnx_name = "xxx.onnx"
model = YOLOv10(model_name)

model.export(format='onnx')

推理命令

./yolov10_cpp best.onnx t.png

python 与C++结果推理不一致

1. 过滤阈值不一致
2. 预处理图像格式不一致（BGR2RGB）