Using Anaconda to estublish a visual environment.

```python pip install ultralytics pip install fastdeploy-gpu-python -f https://link.zhihu.com/?target=https%3A//www.paddlepaddle.org.cn/whl/fastdeploy.html

```

1. Download the

2. (Prepare your own data set);准备自己的数据集:

   (Dataset format requirements);数据集格式需求：

dataset #((Dataset name: e.g. fire)); (数据集名字：例如fire) ├── images ├── train ├── xx.jpg ├── val ├── xx.jpg ├── labels ├── train ├── xx.txt ├── val ├── xx.txt

(Data set partitioning); 数据集划分：

```python

coding:utf-8

import os import random import argparse

parser = argparse.ArgumentParser()

(The address of the xml file, modify it according to your own data. xml is generally stored under Annotations.); xml文件的地址，根据自己的数据进行修改 xml一般存放在Annotations下

parser.addargument('--xmlpath', default='Annotations', type=str, help='input xml label path')

(To divide the data set, select ImageSets/Main under your own data as the address.); 数据集的划分，地址选择自己数据下的ImageSets/Main

parser.addargument('--txtpath', default='ImageSets/Main', type=str, help='output txt label path') opt = parser.parse_args()

trainvalpercent = 1.0 trainpercent = 0.9 xmlfilepath = opt.xmlpath txtsavepath = opt.txtpath total_xml = os.listdir(xmlfilepath) if not os.path.exists(txtsavepath): os.makedirs(txtsavepath)

num = len(totalxml) listindex = range(num) tv = int(num * trainvalpercent) tr = int(tv * trainpercent) trainval = random.sample(list_index, tv) train = random.sample(trainval, tr)

filetrainval = open(txtsavepath + '/trainval.txt', 'w') filetest = open(txtsavepath + '/test.txt', 'w') filetrain = open(txtsavepath + '/train.txt', 'w') fileval = open(txtsavepath + '/val.txt', 'w')

for i in listindex: name = totalxml[i][:-4] + '\n' if i in trainval: filetrainval.write(name) if i in train: filetrain.write(name) else: fileval.write(name) else: filetest.write(name)

filetrainval.close() filetrain.close() fileval.close() filetest.close() ```

.xml to .txt

```python

-- coding: utf-8 --

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

sets = ['train', 'val', 'test'] classes = ["a", "b"] # (Change to your own category); 改成自己的类别 abspath = os.getcwd() print(abspath)

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 convertannotation(imageid): infile = open('/home/trainingai/zyang/yolov5/paperdata/Annotations/%s.xml' % (imageid), encoding='UTF-8') outfile = open('/home/trainingai/zyang/yolov5/paperdata/labels/%s.txt' % (imageid), 'w') tree = ET.parse(infile) 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 difficult = obj.find('Difficult').text cls = obj.find('name').text if cls not in classes or int(difficult) == 1: continue clsid = 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 # (Mark out of bounds correction); 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) outfile.write(str(clsid) + " " + " ".join([str(a) for a in bb]) + '\n')

wd = getcwd() for imageset in sets: if not os.path.exists('/home/trainingai/zyang/yolov5/paperdata/labels/'): os.makedirs('/home/trainingai/zyang/yolov5/paperdata/labels/') imageids = open('/home/trainingai/zyang/yolov5/paperdata/ImageSets/Main/%s.txt' % (imageset)).read().strip().split() listfile = open('/home/trainingai/zyang/yolov5/paperdata/%s.txt' % (imageset), 'w') for imageid in imageids: listfile.write(abspath + '/paperdata/images/%s.jpg\n' % (imageid)) convertannotation(imageid) listfile.close() ```

Add data configuration file

添加数据配置文件

Create a new myvoc.yaml in the yolov5/data folder 在yolov5/data文件夹下新建myvoc.yaml

The content is as follows：

```python

Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]

path: E:\fortestproj\yolov5_ncnn\datasets\fire # dataset root dir train: images/train # train images (relative to 'path') val: images/val # val images (relative to 'path') test: # test images (optional)

Classes

nc: 1 # number of classes names: ['fire'] # class names ```

(Attention please, the train and val parameters here are the paths of the images themselves) (注意，这里的train,val参数是图片本身的路径）

Tips: ./model/yolov5s.yaml, (Just change the category and change it to your own number of categories.); 中改一下类别就可以了，改成你自己的类别数量。

Edit the model’s configuration file:

编辑模型的配置文件：

**Clustering to obtain a priori boxes (optional) (clustering to regenerate anchors takes a long time) *The latest version of yolov5, it will automatically calculate anchors using kmeans* 聚类得出先验框（可选）（聚类重新生成anchors运行时间较长）最新版的yolov5，它会自动kmeans算出anchors

The manual calculation process is as follows: 手动计算过程如下：

Create a new kmeans.py under the dataset folder 在数据集文件夹下新建kmeans.py

```python import numpy as np

def iou(box, clusters): """ Calculates the Intersection over Union (IoU) between a box and k clusters. :param box: tuple or array, shifted to the origin (i. e. width and height) :param clusters: numpy array of shape (k, 2) where k is the number of clusters :return: numpy array of shape (k, 0) where k is the number of clusters """ x = np.minimum(clusters[:, 0], box[0]) y = np.minimum(clusters[:, 1], box[1]) if np.countnonzero(x == 0) > 0 or np.countnonzero(y == 0) > 0: raise ValueError("Box has no area") # (If this error is reported, you can change this line to pass.); 如果报这个错，可以把这行改成pass即可

intersection = x * y
box_area = box[0] * box[1]
cluster_area = clusters[:, 0] * clusters[:, 1]

iou_ = intersection / (box_area + cluster_area - intersection)

return iou_

def avg_iou(boxes, clusters): """ Calculates the average Intersection over Union (IoU) between a numpy array of boxes and k clusters. :param boxes: numpy array of shape (r, 2), where r is the number of rows :param clusters: numpy array of shape (k, 2) where k is the number of clusters :return: average IoU as a single float """ return np.mean([np.max(iou(boxes[i], clusters)) for i in range(boxes.shape[0])])

def translateboxes(boxes): """ Translates all the boxes to the origin. :param boxes: numpy array of shape (r, 4) :return: numpy array of shape (r, 2) """ newboxes = boxes.copy() for row in range(newboxes.shape[0]): newboxes[row][2] = np.abs(newboxes[row][2] - newboxes[row][0]) newboxes[row][3] = np.abs(newboxes[row][3] - newboxes[row][1]) return np.delete(newboxes, [0, 1], axis=1)

def kmeans(boxes, k, dist=np.median): """ Calculates k-means clustering with the Intersection over Union (IoU) metric. :param boxes: numpy array of shape (r, 2), where r is the number of rows :param k: number of clusters :param dist: distance function :return: numpy array of shape (k, 2) """ rows = boxes.shape[0]

distances = np.empty((rows, k))
last_clusters = np.zeros((rows,))

np.random.seed()

# the Forgy method will fail if the whole array contains the same rows
clusters = boxes[np.random.choice(rows, k, replace=False)]

while True:
    for row in range(rows):
        distances[row] = 1 - iou(boxes[row], clusters)

    nearest_clusters = np.argmin(distances, axis=1)

    if (last_clusters == nearest_clusters).all():
        break

    for cluster in range(k):
        clusters[cluster] = dist(boxes[nearest_clusters == cluster], axis=0)

    last_clusters = nearest_clusters

return clusters

if name == 'main': a = np.array([[1, 2, 3, 4], [5, 7, 6, 8]]) print(translate_boxes(a)) ```

Clustering generates the file clauculateanchors.py of new anchors. The code content is as follows: 聚类生成新anchors的文件clauculateanchors.py，代码内容如下：

```python

-- coding: utf-8 --

Find the a priori frame based on the label file

根据标签文件求先验框

import os import numpy as np import xml.etree.cElementTree as et from kmeans import kmeans, avg_iou

FILEROOT = "/home/trainingai/zyang/yolov5/paperdata/" # (root path); 根路径 ANNOTATIONROOT = "Annotations" # (Dataset label folder path); 数据集标签文件夹路径 ANNOTATIONPATH = FILEROOT + ANNOTATIONROOT

ANCHORSTXTPATH = "/home/trainingai/zyang/yolov5/data/anchors.txt"

CLUSTERS = 9 CLASS_NAMES = ['a', 'b']

def loaddata(annodir, classnames): xmlnames = os.listdir(annodir) boxes = [] for xmlname in xmlnames: xmlpth = os.path.join(annodir, xmlname) tree = et.parse(xml_pth)

    width = float(tree.findtext("./size/width"))
    height = float(tree.findtext("./size/height"))

    for obj in tree.findall("./object"):
        cls_name = obj.findtext("name")
        if cls_name in class_names:
            xmin = float(obj.findtext("bndbox/xmin")) / width
            ymin = float(obj.findtext("bndbox/ymin")) / height
            xmax = float(obj.findtext("bndbox/xmax")) / width
            ymax = float(obj.findtext("bndbox/ymax")) / height

            box = [xmax - xmin, ymax - ymin]
            boxes.append(box)
        else:
            continue
return np.array(boxes)

if name == 'main':

anchors_txt = open(ANCHORS_TXT_PATH, "w")

train_boxes = load_data(ANNOTATION_PATH, CLASS_NAMES)
count = 1
best_accuracy = 0
best_anchors = []
best_ratios = []

for i in range(10):      ##### (You can modify it, don’t make it too big, otherwise it will take a long time); 可以修改，不要太大，否则时间很长
    anchors_tmp = []
    clusters = kmeans(train_boxes, k=CLUSTERS)
    idx = clusters[:, 0].argsort()
    clusters = clusters[idx]
    # print(clusters)

    for j in range(CLUSTERS):
        anchor = [round(clusters[j][0] * 640, 2), round(clusters[j][1] * 640, 2)]
        anchors_tmp.append(anchor)
        print(f"Anchors:{anchor}")

    temp_accuracy = avg_iou(train_boxes, clusters) * 100
    print("Train_Accuracy:{:.2f}%".format(temp_accuracy))

    ratios = np.around(clusters[:, 0] / clusters[:, 1], decimals=2).tolist()
    ratios.sort()
    print("Ratios:{}".format(ratios))
    print(20 * "*" + " {} ".format(count) + 20 * "*")

    count += 1

    if temp_accuracy > best_accuracy:
        best_accuracy = temp_accuracy
        best_anchors = anchors_tmp
        best_ratios = ratios

anchors_txt.write("Best Accuracy = " + str(round(best_accuracy, 2)) + '%' + "\r\n")
anchors_txt.write("Best Anchors = " + str(best_anchors) + "\r\n")
anchors_txt.write("Best Ratios = " + str(best_ratios))
anchors_txt.close()

```

After running clauculateanchors.py, a file anchors.txt will be generated, which contains the suggested a priori frame anchors. 运行clauculateanchors.py跑完会生成一个文件 anchors.txt，里面有得出的建议先验框anchors.

Modify the path and parameters in train.py

修改train.py中的路径及参数

The parameters are explained as follows: Epochs: refers to how many times the entire data set will be iterated during the training process. If the graphics card is not good, you can adjust it smaller. Batch-size: How many pictures are viewed at one time before the weights are updated. Gradient descent mini-batch. If the graphics card is not good, you can adjust it smaller. cfg: Configuration file that stores the model structure data: files that store training and test data img-size: Enter the image width and height. If the graphics card is not good, you can adjust it smaller. rect: perform rectangular training resume: Resume the recently saved model to start training nosave: only save the final checkpoint notest: only test the last epoch evolve: evolve hyperparameters bucket: gsutil bucket cache-images: cache images to speed up training weights: weight file path name: Rename results.txt to results_name.txt device: cuda device, i.e. 0 or 0,1,2,3 or cpu adam: Use adam optimization multi-scale: multi-scale training, img-size +/- 50% single-cls: single-category training set

参数解释如下： epochs：指的就是训练过程中整个数据集将被迭代多少次,显卡不行你就调小点。 batch-size：一次看完多少张图片才进行权重更新，梯度下降的mini-batch,显卡不行你就调小点。 cfg：存储模型结构的配置文件 data：存储训练、测试数据的文件 img-size：输入图片宽高,显卡不行你就调小点。 rect：进行矩形训练 resume：恢复最近保存的模型开始训练 nosave：仅保存最终checkpoint notest：仅测试最后的epoch evolve：进化超参数 bucket：gsutil bucket cache-images：缓存图像以加快训练速度 weights：权重文件路径 name： 重命名results.txt to results_name.txt device：cuda device, i.e. 0 or 0,1,2,3 or cpu adam：使用adam优化 multi-scale：多尺度训练，img-size +/- 50% single-cls：单类别的训练集

Reference:

1. https://blog.csdn.net/qq_36756866/article/details/109111065
2. https://blog.csdn.net/weixin70251903/article/details/126813751?utmmedium=distribute.pcrelevant.none-task-blog-2~default~baidujsbaidulandingword~default-4-126813751-blog-122874005.235^v38^pcrelevantsortbase1&spm=1001.2101.3001.4242.3&utmrelevant_index=7
3. https://blog.csdn.net/qq52859223/article/details/123701798?utmmedium=distribute.pcrelevant.none-task-blog-2~default~baidujsbaidulandingword~default-0-123701798-blog-109111065.235^v38^pcrelevantsortbase1&spm=1001.2101.3001.4242.1&utmrelevant_index=1
4. https://zhuanlan.zhihu.com/p/501798155