pytorch篇---yolov5-权重转换（pt--＞onnx--＞rknn）

yolov5训练自己的数据集

yolov5官方代码： 选择 tag v5.0
https://github.com/ultralytics/yolov5

1.制作数据集

voc 数据集变成yolov5所需的格式

格式转换

# -*- coding: utf-8 -*-

import xml.etree.ElementTree as ET
import os
from os import getcwd

sets = ['train', 'val', 'test']
classes = ['TrashBin', 'overflowed', 'overground']  #类别数目
abs_path = os.getcwd()


def convert(size, box):
    dw = 1. / (size[0])
    dh = 1. / (size[1])
    x = (box[0] + box[1]) / 2.0 - 1
    y = (box[2] + box[3]) / 2.0 - 1
    w = box[1] - box[0]
    h = box[3] - box[2]
    x = x * dw
    w = w * dw
    y = y * dh
    h = h * dh
    return x, y, w, h


def convert_annotation(image_id):
    in_file = open('D:/second_project/data/all_dataset2021_yolov5/Annotations/%s.xml' % (image_id), encoding='utf-8')
    out_file = open('D:/second_project/data/all_dataset2021_yolov5/labels/%s.txt' % (image_id), mode='w', encoding='utf-8')
    tree = ET.parse(in_file)
    root = tree.getroot()
    size = root.find('size')
    w = int(size.find('width').text)
    h = int(size.find('height').text)
    for obj in root.iter('object'):
        difficult = obj.find('difficult').text
        cls = obj.find('name').text
        if cls not in classes or int(difficult) == 1:
            continue
        cls_id = classes.index(cls)
        xmlbox = obj.find('bndbox')
        b = (float(xmlbox.find('xmin').text), float(xmlbox.find('xmax').text), float(xmlbox.find('ymin').text),
             float(xmlbox.find('ymax').text))
        b1, b2, b3, b4 = b
        # 标注越界修正
        if b2 > w:
            b2 = w
        if b4 > h:
            b4 = h
        b = (b1, b2, b3, b4)
        bb = convert((w, h), b)
        out_file.write(str(cls_id) + " " + " ".join([str(a) for a in bb]) + '\n')


wd = getcwd()
for image_set in sets:
    if not os.path.exists('D:/second_project/data/all_dataset2021_yolov5/labels'):
        os.makedirs('D:/second_project/data/all_dataset2021_yolov5/labels')
    image_ids = open('D:/second_project/data/all_dataset2021_yolov5/ImageSets/Main/%s.txt' % (image_set), encoding='utf-8').read().strip().split()
    list_file = open('%s.txt' % (image_set), mode='w', encoding='utf-8')
    for image_id in image_ids:
        list_file.write('D:/second_project/data/all_dataset2021_yolov5/images/%s.jpg\n' % (image_id))
        convert_annotation(image_id)
    list_file.close()

如图：labels是新生成的

2. 训练数据 train.py(官方的)

修改的地方如下：
2.1

2.2

2.3 若选择yolov5s.pt 作为初始化权重，需要将 nc 改为自己的标签类别个数
若送进网络的大小不是（640， 640）需要修改锚框

训练完成后：

3. 将pt权重转换成onnx

参考代码：https://github.com/soloIife/yolov5_for_rknn
export_no_focus.py :

python models/export_no_focus.py --weights "xxx.pt"

4. onnx转换成rknn

yolov5_for_rknn-master/yolov5_original 下的
onnx2rknn.py

python onnx2rknn.py --weights/xxx.onnx  --precompile  --original

5. rknn推理

yolov5_for_rknn-master/yolov5_original 下的
rknn_detect.py

python rknn_detect.py