Source code analysis 2 Model conversion export.py

基于最新yolov5-v6.0

​ ​1.重点​​

​ ​2.相关函数​​

1.重点

General rules of use：

登录后复制

      
      

       
       python export.py --data " " --weights " " --imgsz 512 --simplify --include "onnx"
      
      
      
      

       
       
1.
      
      

Which of the three formats you want to use needs to download the corresponding package：

1. torchscript There is no need to download the corresponding package 有Torch就可以
2. onnx: pip install onnx
3. coreml: pip install coremltools

2.相关函数

parse_opt():

登录后复制

      
      

       
       def parse_opt():
       
       
    """
       
       
    data: 数据集目录 默认=ROOT / 'data/coco128.yaml'
       
       
    weights：Weights file directory 默认=ROOT / 'yolov5s.pt'
       
       
    img-size: Enter a picture of the modelsize=(height, width) 默认=[640, 640]
       
       
    batch-size: batch大小 默认=1
       
       
    device: Model running equipment cuda device, i.e. 0 or 0,1,2,3 or cpu 默认=cpu
       
       
    include: 要将ptWhat format the file is converted to Can be single primitive or list 默认=['torchscript', 'onnx', 'coreml']
       
       
    half: 是否使用半精度FP16export转换 默认=False
       
       
    inplace: 是否设置 YOLOv5 Detect() inplace=True  默认=False
       
       
    train: 是否开启model.train() mode 默认=True  coremlconversion must beTrue
       
       
    optimize: TorchScript转化参数 Whether to perform mobile optimization  默认=False
       
       
    int8： 支持CoreML/TF INT8 量化 不支持ONNX
       
       
    dynamic: ONNX转换参数  dynamic_axes  ONNXWhether the transformation is to do batch variables  默认=False
       
       
    simplify: ONNX转换参数 是否简化onnx模型  默认=False
       
       
    opset: ONNX转换参数 设置ONNX版本  默认=13
       
       
    topk-per-class: TF.jsEach category is reserved 默认=100
       
       
    topk-all: TF.js  Topk为所有class保留
       
       
    iou-thres: TF.js IoU threshold   default=0.45
       
       
    conf-thres: TF.js  confidence threshold  default=0.25
       
       
    include:     The version that needs to be exported  default=['torchscript', 'onnx'],
       
       
    """
      
      
      
      

       
       
1.
       
       
2.
       
       
3.
       
       
4.
       
       
5.
       
       
6.
       
       
7.
       
       
8.
       
       
9.
       
       
10.
       
       
11.
       
       
12.
       
       
13.
       
       
14.
       
       
15.
       
       
16.
       
       
17.
       
       
18.
       
       
19.
       
       
20.
       
       
21.
       
       
22.
      
      

def run(data=ROOT / 'data/coco128.yaml', # 'dataset.yaml path'

登录后复制

      
      

       
       def run(data=ROOT / 'data/coco128.yaml',  # 'dataset.yaml path'
       
       
        weights=ROOT / 'yolov5s.pt',  # weights path
       
       
        imgsz=(640, 640),  # image (height, width)
       
       
        batch_size=1,  # batch size
       
       
        device='cpu',  # cuda device, i.e. 0 or 0,1,2,3 or cpu
       
       
        include=('torchscript', 'onnx', 'coreml'),  # include formats
       
       
        half=False,  # FP16 half-precision export
       
       
        inplace=False,  # set YOLOv5 Detect() inplace=True
       
       
        train=False,  # model.train() mode
       
       
        optimize=False,  # TorchScript: optimize for mobile
       
       
        int8=False,  # CoreML/TF INT8 quantization
       
       
        dynamic=False,  # ONNX/TF: dynamic axes
       
       
        simplify=False,  # ONNX: simplify model
       
       
        opset=12,  # ONNX: opset version
       
       
        topk_per_class=100,  # TF.js NMS: topk per class to keep
       
       
        topk_all=100,  # TF.js NMS: topk for all classes to keep
       
       
        iou_thres=0.45,  # TF.js NMS: IoU threshold
       
       
        conf_thres=0.25  # TF.js NMS: confidence threshold
       
       
        ):
      
      
      
      

       
       
1.
       
       
2.
       
       
3.
       
       
4.
       
       
5.
       
       
6.
       
       
7.
       
       
8.
       
       
9.
       
       
10.
       
       
11.
       
       
12.
       
       
13.
       
       
14.
       
       
15.
       
       
16.
       
       
17.
       
       
18.
       
       
19.