pytorch框架yolov3算法训练自己数据集 win7~win10

1. 环境搭建

首先需要安装的工具pytorch，这是一个从torch改进过来的框架，进入官网的网站，选择自己的电脑系统，下载方式，语言，还有对应的CUDA，此时会给出一行pip命令，复制命令，打开cmd命令行，粘贴，同时加上国内源提高下载速度，等待2个钟头左右，pytorch安装完毕。
不同的pytorch有不同的CUDA和CUDNN的版本。CUDA和CUDNN英伟达出品的并行计算架构，简单通俗的说就是可以让GPU代替CPU的计算，当然由于是并发，所以处理大量运算的时候速度更快，在下载CUDA的时候一定要注意是否和之前pytorch下载的时候选择的CUDA版本一致，下载地址就在英伟达的官方网站首页。Cudnn是一个加速库，面向神经网络，同样可以在英伟达的官网找到，也要注意支持的CUDA版本。
在全部下载完毕后，输出代码torch.cuda.device(0)和torch.cuda.is.available()测试GPU是否能正常使用，如果不行说明CUDA没有安装好，或者显卡不支持。
pytorch安装连接 https://pytorch.org/

2. 数据集构建

1. xml文件生成需要Labelimg软件

在Windows下使用LabelImg软件进行标注，快捷键：

Ctrl + u 加载目录中的所有图像，鼠标点击Open dir同功能
Ctrl + r 更改默认注释目标目录(xml文件保存的地址)
Ctrl + s 保存
Ctrl + d 复制当前标签和矩形框 space 将当前图像标记为已验证 w
创建一个矩形框
d 下一张图片
a 上一张图片
del 删除选定的矩形框
Ctrl++ 放大
Ctrl-- 缩小
↑→↓← 键盘箭头移动选定的矩形框

2. VOC2007 数据集格式

-data
---- VOCdevkit2007
---- VOC2007
---- Annotations (标签XML文件，用对应的图片处理工具人工生成的)
---- ImageSets (生成的方法是用sh或者MATLAB语言生成)
---- ---- test.txt
---- ---- train.txt
---- ---- trainval.txt
---- ---- val.txt
- JPEGImages(原始文件)
- labels (xml文件对应的txt文件)

imageSets下面的文件可以由makeTxt.py生成

import os
import random

trainval_percent = 0.1
train_percent = 0.9
xmlfilepath = 'data/Annotations'
txtsavepath = 'data/ImageSets'
total_xml = os.listdir(xmlfilepath)

num = len(total_xml)
list = range(num)
tv = int(num * trainval_percent)
tr = int(tv * train_percent)
trainval = random.sample(list, tv)
train = random.sample(trainval, tr)

ftrainval = open('data/ImageSets/trainval.txt', 'w')
ftest = open('data/ImageSets/test.txt', 'w')
ftrain = open('data/ImageSets/train.txt', 'w')
fval = open('data/ImageSets/val.txt', 'w')

for i in list:
    name = total_xml[i][:-4] + '\n'
    if i in trainval:
        ftrainval.write(name)
        if i in train:
            ftest.write(name)
        else:
            fval.write(name)
    else:
        ftrain.write(name)

ftrainval.close()
ftrain.close()
fval.close()
ftest.close()

xml对应的txt文件由voc_label.py生成 代码如下

import xml.etree.ElementTree as ET
import pickle
import os
from os import listdir, getcwd
from os.path import join

sets = ['train', 'test', 'val']

classes = ["face","face_mask"]  # 我们只是检测细胞，因此只有一个类别


def convert(size, box):
    dw = 1. / size[0]
    dh = 1. / size[1]
    x = (box[0] + box[1]) / 2.0
    y = (box[2] + box[3]) / 2.0
    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('data/Annotations/%s.xml' % (image_id))
    out_file = open('data/labels/%s.txt' % (image_id), 'w')
    tree = ET.parse(in_file)
    root = tree.getroot()
    size = root.find('size')
    height=root.find('height')
    w=1024
    h=1024
    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))
        bb = convert((w, h), b)
        out_file.write(str(cls_id) + " " + " ".join([str(a) for a in bb]) + '\n')
       # out_file.write(str(cls_id) + " " + " ".join([str(a) for a in b]) + '\n')


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

此时为了避免错误检查一下xml文件，是否有size=0的情况，可以通关python进行批量删除，代码如下：

import xml.etree.ElementTree as ET
import pickle
import os
from os import listdir, getcwd
from os.path import join
def convert_annotation(image_id):
    in_file = open('data/Annotations/%s.xml' % (image_id))
    path='data/Annotations/%s.xml' % (image_id)
    path2='data/Annotations/%s.xml' % (image_id)
    tree = ET.parse(in_file)
    root = tree.getroot()
    size = root.find('size')
    if(size is None):
        print(image_id)
        in_file.close()
        os.remove(path)
    else:
        w=int(size.find('width').text)
        h = int(size.find('height').text)
        if(w==0 or h==0):
            print(image_id)
            in_file.close()
            os.remove(path)      
image_ids =[x for x in range(100,6366)]
for image_id in image_ids:
    convert_annotation(image_id)

若有size=0的情况，可重新执行上面两段代码

3. 创建*.names

其中保存的是你的所有的类别，每行一个类别，如data/mask.names：
face face_mask

4. 更新data/mask.data

classes = 2 # 改成你的数据集的类别个数
train = data/train.txt # 通过voc_label.py文件生成的txt文件
valid = data/test.txt # 通过voc_label.py文件生成的txt文件
names = data/mask.names # 记录类别
backup = backup/ # 记录checkpoint存放位置
eval = coco # 选择map计算方式

5. 更新cfg文件，修改类别相关信息

[net]
# Testing
batch=1
subdivisions=1
# Training
# batch=64
# subdivisions=2
width=416
height=416
channels=3
momentum=0.9
decay=0.0005
angle=0
saturation = 1.5
exposure = 1.5
hue=.1

learning_rate=0.001
burn_in=1000
max_batches = 500200
policy=steps
steps=400000,450000
scales=.1,.1

[convolutional]
batch_normalize=1
filters=16
size=3
stride=1
pad=1
activation=leaky

[maxpool]
size=2
stride=2

[convolutional]
batch_normalize=1
filters=32
size=3
stride=1
pad=1
activation=leaky

[maxpool]
size=2
stride=2

[convolutional]
batch_normalize=1
filters=64
size=3
stride=1
pad=1
activation=leaky

[maxpool]
size=2
stride=2

[convolutional]
batch_normalize=1
filters=128
size=3
stride=1
pad=1
activation=leaky

[maxpool]
size=2
stride=2

[convolutional]
batch_normalize=1
filters=256
size=3
stride=1
pad=1
activation=leaky

[maxpool]
size=2
stride=2

[convolutional]
batch_normalize=1
filters=512
size=3
stride=1
pad=1
activation=leaky

[maxpool]
size=2
stride=1

[convolutional]
batch_normalize=1
filters=1024
size=3
stride=1
pad=1
activation=leaky

###########

[convolutional]
batch_normalize=1
filters=256
size=1
stride=1
pad=1
activation=leaky

[convolutional]
batch_normalize=1
filters=512
size=3
stride=1
pad=1
activation=leaky

[convolutional]
size=1
stride=1
pad=1
filters=21
activation=linear



[yolo]
mask = 3,4,5
anchors = 10,14,  23,27,  37,58,  81,82,  135,169,  344,319
classes=2
num=6
jitter=.3
ignore_thresh = .7
truth_thresh = 1
random=1

[route]
layers = -4

[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky

[upsample]
stride=2

[route]
layers = -1, 8

[convolutional]
batch_normalize=1
filters=256
size=3
stride=1
pad=1
activation=leaky

[convolutional]
size=1
stride=1
pad=1
filters=21
activation=linear

[yolo]
mask = 0,1,2
anchors = 10,14,  23,27,  37,58,  81,82,  135,169,  344,319
classes=2
num=6
jitter=.3
ignore_thresh = .7
truth_thresh = 1
random=1

简单来说tiny中就是把classes改为需要测的分类，此处为2 classes上一层[convolutional]中的filters改为3*（5+classes）

3. 训练模型

首先打开cmd命令,进入项目的路径。使用python train.py运行训练代码,效果如图所示:

4. 测试图片和视频.

python detect.py --weights weights/best.pt

import argparse
import time
from sys import platform

from models import *
from project.datasets import *
from project.utils import *


def detect(
        cfg,
        data_cfg,
        weights,
        images='data/samples',  # input folder
        output='output',  # output folder
        fourcc='mp4v',
        img_size=416,
        conf_thres=0.5,
        nms_thres=0.5,
        save_txt=False,
        save_images=True,
        webcam=False
):
    device = torch_utils.select_device()
    if os.path.exists(output):
        shutil.rmtree(output)  # delete output folder
    os.makedirs(output)  # make new output folder

    # Initialize model
    if ONNX_EXPORT:
        s = (320, 192)  # onnx model image size (height, width)
        model = Darknet(cfg, s)
    else:
        model = Darknet(cfg, img_size)

    # Load weights
    if weights.endswith('.pt'):  # pytorch format
        model.load_state_dict(torch.load(weights, map_location=device)['model'])
    else:  # darknet format
        _ = load_darknet_weights(model, weights)

    # Fuse Conv2d + BatchNorm2d layers
    model.fuse()

    # Eval mode
    model.to(device).eval()

    if ONNX_EXPORT:
        img = torch.zeros((1, 3, s[0], s[1]))
        torch.onnx.export(model, img, 'weights/export.onnx', verbose=True)
        return

    # Set Dataloader
    vid_path, vid_writer =None,None
    if webcam:
        save_images = True
        dataloader = LoadWebcam(img_size=img_size)
    else:
        dataloader = LoadImages(images, img_size=img_size)

    # Get classes and colors
    classes = load_classes(parse_data_cfg(data_cfg)['names'])
    colors = [[random.randint(0, 255) for _ in range(3)] for _ in range(len(classes))]
    for i, (path, img, im0, vid_cap) in enumerate(dataloader):
        t = time.time()
        save_path = str(Path(output) / Path(path).name)

        # Get detections
        img = torch.from_numpy(img).unsqueeze(0).to(device)
        pred, _ = model(img)
        det = non_max_suppression(pred, conf_thres, nms_thres)[0]

        if det is not None and len(det) > 0:
            # Rescale boxes from 416 to true image size
            det[:, :4] = scale_coords(img.shape[2:], det[:, :4], im0.shape).round()

            # Print results to screen
            print('%gx%g ' % img.shape[2:], end='')  # print image size
            for c in det[:, -1].unique():
                n = (det[:, -1] == c).sum()
                print('%g %ss' % (n, classes[int(c)]), end=', ')

            # Draw bounding boxes and labels of detections
            for *xyxy, conf, cls_conf, cls in det:
                if save_txt:  # Write to file
                    with open(save_path + '.txt', 'a') as file:
                        file.write(('%g ' * 6 + '\n') % (*xyxy, cls, conf))

                # Add bbox to the image
                label = '%s %.2f' % (classes[int(cls)], conf)
                plot_one_box(xyxy, im0, label=label, color=colors[int(cls)])

        print('Done. (%.3fs)' % (time.time() - t))

        if webcam:  # Show live webcam
            cv2.imshow(weights, im0)
           
        if save_images:  # Save image with detections
            if dataloader.mode == 'images':
                cv2.imwrite(save_path, im0)
            else:
                if vid_path != save_path:  # new video
                    vid_path = save_path
                    if isinstance(vid_writer, cv2.VideoWriter):
                        vid_writer.release()  # release previous video writer

                    fps = vid_cap.get(cv2.CAP_PROP_FPS)
                    width = int(vid_cap.get(cv2.CAP_PROP_FRAME_WIDTH))
                    height = int(vid_cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
                    vid_writer = cv2.VideoWriter(save_path, cv2.VideoWriter_fourcc(*fourcc), fps, (width, height))
                vid_writer.write(im0)

    if save_images:
        print('Results saved to %s' % os.getcwd() + os.sep + output)
        if platform == 'darwin':  # macos
            os.system('open ' + output + ' ' + save_path)


if __name__ == '__main__':
    parser = argparse.ArgumentParser()
    parser.add_argument('--cfg', type=str, default='cfg/yolov3-tiny.cfg', help='cfg file path')
    parser.add_argument('--data-cfg', type=str, default='data/mask.data', help='coco.data file path')
    parser.add_argument('--weights', type=str, default='weights/latest.pt', help='path to weights file')
    parser.add_argument('--images', type=str, default='data/samples', help='path to images')
    parser.add_argument('--img-size', type=int, default=416, help='inference size (pixels)')
    parser.add_argument('--conf-thres', type=float, default=0.5, help='object confidence threshold')
    parser.add_argument('--nms-thres', type=float, default=0.5, help='iou threshold for non-maximum suppression')
    parser.add_argument('--fourcc', type=str, default='mp4v', help='specifies the fourcc code for output video encoding (make sure ffmpeg supports specified fourcc codec)')
    parser.add_argument('--output', type=str, default='output',help='specifies the output path for images and videos')
    opt = parser.parse_args()
    print(opt)

    with torch.no_grad():
        detect(
            opt.cfg,
            opt.data_cfg,
            opt.weights,
            images=opt.images,
            img_size=opt.img_size,
            conf_thres=opt.conf_thres,
            nms_thres=opt.nms_thres,
            fourcc=opt.fourcc,
            output=opt.output
        )

5. 测试摄像头

把detect.py中的webcam改为False。此时会报错NoneType，打开datasets.py文件

import glob
import math
import os
import random
import shutil
from pathlib import Path

import cv2
import numpy as np
import torch
from torch.utils.data import Dataset
from tqdm import tqdm

from project.utils import xyxy2xywh


class LoadImages:  # for inference
    def __init__(self, path, img_size=416):
        self.height = img_size
        img_formats = ['.jpg', '.jpeg', '.png', '.tif']
        vid_formats = ['.mov', '.avi', '.mp4']

        files = []
        if os.path.isdir(path):
            files = sorted(glob.glob('%s/*.*' % path))
        elif os.path.isfile(path):
            files = [path]

        images = [x for x in files if os.path.splitext(x)[-1].lower() in img_formats]
        videos = [x for x in files if os.path.splitext(x)[-1].lower() in vid_formats]
        nI, nV = len(images), len(videos)

        self.files = images + videos
        self.nF = nI + nV  # number of files
        self.video_flag = [False] * nI + [True] * nV
        self.mode = 'images'
        if any(videos):
            self.new_video(videos[0])  # new video
        else:
            self.cap = None
        assert self.nF > 0, 'No images or videos found in ' + path

    def __iter__(self):
        self.count = 0
        return self

    def __next__(self):
        if self.count == self.nF:
            raise StopIteration
        path = self.files[self.count]

        if self.video_flag[self.count]:
            # Read video
            self.mode = 'video'
            ret_val, img0 = self.cap.read()
            if not ret_val:
                self.count += 1
                self.cap.release()
                if self.count == self.nF:  # last video
                    raise StopIteration
                else:
                    path = self.files[self.count]
                    self.new_video(path)
                    ret_val, img0 = self.cap.read()

            self.frame += 1
            print('video %g/%g (%g/%g) %s: ' % (self.count + 1, self.nF, self.frame, self.nframes, path), end='')

        else:
            # Read image
            self.count += 1
            img0 = cv2.imread(path)  # BGR
            assert img0 is not None, 'File Not Found ' + path
            print('image %g/%g %s: ' % (self.count, self.nF, path), end='')

        # Padded resize
        img, _, _, _ = letterbox(img0, new_shape=self.height)

        # Normalize RGB
        img = img[:, :, ::-1].transpose(2, 0, 1)  # BGR to RGB
        img = np.ascontiguousarray(img, dtype=np.float32)  # uint8 to float32
        img /= 255.0  # 0 - 255 to 0.0 - 1.0

        # cv2.imwrite(path + '.letterbox.jpg', 255 * img.transpose((1, 2, 0))[:, :, ::-1])  # save letterbox image
        return path, img, img0, self.cap

    def new_video(self, path):
        self.frame = 0
        self.cap = cv2.VideoCapture(path)#path
        self.nframes = int(self.cap.get(cv2.CAP_PROP_FRAME_COUNT))

    def __len__(self):
        return self.nF  # number of files


class LoadWebcam:  # for inference
    def __init__(self, img_size=416):
        vid_formats = ['.mov', '.avi', '.mp4']#64
        files = []#64

        self.cam = cv2.VideoCapture(0)
        self.height = img_size
        videos = [x for x in files if os.path.splitext(x)[-1].lower() in vid_formats]#64
        self.mode = 'images'#64
        if any(videos):#64
            self.new_video(videos[0])  # new video#64
        else:#64
            self.cap = None#64
    def __iter__(self):
        self.count = -1
        return self

    def __next__(self):
        self.count += 1
        if cv2.waitKey(1) == 27:  # esc to quit
            cv2.destroyAllWindows()
            raise StopIteration

        # Read image

        ret_val, img0 = self.cam.read()
        assert ret_val, 'Webcam Error'
        img_path = 'webcam_%g.jpg' % self.count
        img0 = cv2.flip(img0, 1)  # flip left-right
        print('webcam %g: ' % self.count, end='')

        # Padded resize
        img, _, _, _ = letterbox(img0, new_shape=self.height)

        # Normalize RGB
        img = img[:, :, ::-1].transpose(2, 0, 1)  # BGR to RGB
        img = np.ascontiguousarray(img, dtype=np.float32)  # uint8 to float32
        img /= 255.0  # 0 - 255 to 0.0 - 1.0

        return img_path, img, img0, None

    def __len__(self):
        return 0


class LoadImagesAndLabels(Dataset):  # for training/testing
    def __init__(self, path, img_size=416, batch_size=16, augment=False, rect=True, image_weights=False):
        with open(path, 'r') as f:
            img_files = f.read().splitlines()
            self.img_files = list(filter(lambda x: len(x) > 0, img_files))

        n = len(self.img_files)
        bi = np.floor(np.arange(n) / batch_size).astype(np.int)  # batch index
        nb = bi[-1] + 1  # number of batches
        assert n > 0, 'No images found in %s' % path

        self.n = n
        self.batch = bi  # batch index of image
        self.img_size = img_size
        self.augment = augment
        self.image_weights = image_weights
        self.rect = False if image_weights else rect
        self.label_files = [x.replace('images', 'labels').
                                replace('.jpeg', '.txt').
                                replace('.jpg', '.txt').
                                replace('.bmp', '.txt').
                                replace('.png', '.txt') for x in self.img_files]

        # Rectangular Training  https://github.com/ultralytics/yolov3/issues/232
        if self.rect:
            from PIL import Image

            # Read image shapes
            #sp = 'data' + os.sep + path.replace('.txt', '.shapes').split(os.sep)[-1]  # shapefile path
            #sp = os.sep + path.replace('.txt', '.shapes').split(os.sep)[-1]  # shapefile path
            sp=path. replace('.txt','shapes'). split(os.sep)[-1]
            if os.path.exists(sp):  # read existing shapefile
                with open(sp, 'r') as f:
                    s = np.array([x.split() for x in f.read().splitlines()], dtype=np.float32)
                assert len(s) == n, 'Shapefile out of sync, please delete %s and rerun' % sp
            else:  # no shapefile, so read shape using PIL and write shapefile for next time (faster)
                s = np.array([Image.open(f).size for f in tqdm(self.img_files, desc='Reading image shapes')])
                np.savetxt(sp, s, fmt='%g')

            # Sort by aspect ratio
            ar = s[:, 1] / s[:, 0]  # aspect ratio
            i = ar.argsort()
            ar = ar[i]
            self.img_files = [self.img_files[i] for i in i]
            self.label_files = [self.label_files[i] for i in i]

            # Set training image shapes
            shapes = [[1, 1]] * nb
            for i in range(nb):
                ari = ar[bi == i]
                mini, maxi = ari.min(), ari.max()
                if maxi < 1:
                    shapes[i] = [maxi, 1]
                elif mini > 1:
                    shapes[i] = [1, 1 / mini]

            self.batch_shapes = np.ceil(np.array(shapes) * img_size / 32.).astype(np.int) * 32

        # Preload labels (required for weighted CE training)
        self.imgs = [None] * n
        self.labels = [np.zeros((0, 5))] * n
        iter = tqdm(self.label_files, desc='Reading labels') if n > 1000 else self.label_files
        for i, file in enumerate(iter):
            try:
                with open(file, 'r') as f:
                    l = np.array([x.split() for x in f.read().splitlines()], dtype=np.float32)
                    if l.shape[0]:
                        assert l.shape[1] == 5, '> 5 label columns: %s' % file
                        assert (l >= 0).all(), 'negative labels: %s' % file
                        assert (l[:, 1:] <= 1).all(), 'non-normalized or out of bounds coordinate labels: %s' % file
                        self.labels[i] = l
            except:
                pass  # print('Warning: missing labels for %s' % self.img_files[i])  # missing label file
        assert len(np.concatenate(self.labels, 0)) > 0, 'No labels found. Incorrect label paths provided.'

    def __len__(self):
        return len(self.img_files)

  #  def __iter__(self):
   #     self.count = -1
   #     print('ran dataset iter')
    #    self.shuffled_vector = np.random.permutation(self.nF) if self.augment else np.arange(self.nF)
     #   return self

    def __getitem__(self, index):
        if self.image_weights:
            index = self.indices[index]

        img_path = self.img_files[index]
        label_path = self.label_files[index]

        # Load image
        img = self.imgs[index]
        if img is None:
            img = cv2.imread(img_path)  # BGR
            assert img is not None, 'File Not Found ' + img_path
            if self.n < 1001:
                self.imgs[index] = img  # cache image into memory

        # Augment colorspace
        augment_hsv = True
        if self.augment and augment_hsv:
            # SV augmentation by 50%
            fraction = 0.50  # must be < 1.0
            img_hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)  # hue, sat, val
            S = img_hsv[:, :, 1].astype(np.float32)  # saturation
            V = img_hsv[:, :, 2].astype(np.float32)  # value

            a = (random.random() * 2 - 1) * fraction + 1
            b = (random.random() * 2 - 1) * fraction + 1
            S *= a
            V *= b

            img_hsv[:, :, 1] = S if a < 1 else S.clip(None, 255)
            img_hsv[:, :, 2] = V if b < 1 else V.clip(None, 255)
            cv2.cvtColor(img_hsv, cv2.COLOR_HSV2BGR, dst=img)

        # Letterbox
        h, w, _ = img.shape
        if self.rect:
            shape = self.batch_shapes[self.batch[index]]
            img, ratio, padw, padh = letterbox(img, new_shape=shape, mode='rect')
        else:
            shape = self.img_size
            img, ratio, padw, padh = letterbox(img, new_shape=shape, mode='square')

        # Load labels
        labels = []
        if os.path.isfile(label_path):
            # with open(label_path, 'r') as f:
            #     x = np.array([x.split() for x in f.read().splitlines()], dtype=np.float32)
            x = self.labels[index]
            if x.size > 0:
                # Normalized xywh to pixel xyxy format
                labels = x.copy()
                labels[:, 1] = ratio * w * (x[:, 1] - x[:, 3] / 2) + padw
                labels[:, 2] = ratio * h * (x[:, 2] - x[:, 4] / 2) + padh
                labels[:, 3] = ratio * w * (x[:, 1] + x[:, 3] / 2) + padw
                labels[:, 4] = ratio * h * (x[:, 2] + x[:, 4] / 2) + padh

        # Augment image and labels
        if self.augment:
            img, labels = random_affine(img, labels, degrees=(-5, 5), translate=(0.10, 0.10), scale=(0.90, 1.10))

        nL = len(labels)  # number of labels
        if nL:
            # convert xyxy to xywh
            labels[:, 1:5] = xyxy2xywh(labels[:, 1:5])

            # Normalize coordinates 0 - 1
            labels[:, [2, 4]] /= img.shape[0]  # height
            labels[:, [1, 3]] /= img.shape[1]  # width

        if self.augment:
            # random left-right flip
            lr_flip = True
            if lr_flip and random.random() > 0.5:
                img = np.fliplr(img)
                if nL:
                    labels[:, 1] = 1 - labels[:, 1]

            # random up-down flip
            ud_flip = False
            if ud_flip and random.random() > 0.5:
                img = np.flipud(img)
                if nL:
                    labels[:, 2] = 1 - labels[:, 2]

        labels_out = torch.zeros((nL, 6))
        if nL:
            labels_out[:, 1:] = torch.from_numpy(labels)

        # Normalize
        img = img[:, :, ::-1].transpose(2, 0, 1)  # BGR to RGB, to 3x416x416
        img = np.ascontiguousarray(img, dtype=np.float32)  # uint8 to float32
        img /= 255.0  # 0 - 255 to 0.0 - 1.0

        return torch.from_numpy(img), labels_out, img_path, (h, w)

    @staticmethod
    def collate_fn(batch):
        img, label, path, hw = list(zip(*batch))  # transposed
        for i, l in enumerate(label):
            l[:, 0] = i  # add target image index for build_targets()
        return torch.stack(img, 0), torch.cat(label, 0), path, hw


def letterbox(img, new_shape=416, color=(127.5, 127.5, 127.5), mode='auto'):
    # Resize a rectangular image to a 32 pixel multiple rectangle
    # https://github.com/ultralytics/yolov3/issues/232
    shape = img.shape[:2]  # current shape [height, width]
    if isinstance(new_shape, int):
        ratio = float(new_shape) / max(shape)
    else:
        ratio = max(new_shape) / max(shape)  # ratio  = new / old
    new_unpad = (int(round(shape[1] * ratio)), int(round(shape[0] * ratio)))

    # Compute padding https://github.com/ultralytics/yolov3/issues/232
    if mode is 'auto':  # minimum rectangle
        dw = np.mod(new_shape - new_unpad[0], 32) / 2  # width padding
        dh = np.mod(new_shape - new_unpad[1], 32) / 2  # height padding
    elif mode is 'square':  # square
        dw = (new_shape - new_unpad[0]) / 2  # width padding
        dh = (new_shape - new_unpad[1]) / 2  # height padding
    elif mode is 'rect':  # square
        dw = (new_shape[1] - new_unpad[0]) / 2  # width padding
        dh = (new_shape[0] - new_unpad[1]) / 2  # height padding

    top, bottom = int(round(dh - 0.1)), int(round(dh + 0.1))
    left, right = int(round(dw - 0.1)), int(round(dw + 0.1))
    img = cv2.resize(img, new_unpad, interpolation=cv2.INTER_LINEAR)  # resized, no border
    img = cv2.copyMakeBorder(img, top, bottom, left, right, cv2.BORDER_CONSTANT, value=color)  # padded square
    return img, ratio, dw, dh


def random_affine(img, targets=(), degrees=(-10, 10), translate=(.1, .1), scale=(.9, 1.1), shear=(-2, 2),
                  borderValue=(127.5, 127.5, 127.5)):
    # torchvision.transforms.RandomAffine(degrees=(-10, 10), translate=(.1, .1), scale=(.9, 1.1), shear=(-10, 10))
    # https://medium.com/uruvideo/dataset-augmentation-with-random-homographies-a8f4b44830d4

    if targets is None:
        targets = []
    border = 0  # width of added border (optional)
    height = img.shape[0] + border * 2
    width = img.shape[1] + border * 2

    # Rotation and Scale
    R = np.eye(3)
    a = random.random() * (degrees[1] - degrees[0]) + degrees[0]
    # a += random.choice([-180, -90, 0, 90])  # 90deg rotations added to small rotations
    s = random.random() * (scale[1] - scale[0]) + scale[0]
    R[:2] = cv2.getRotationMatrix2D(angle=a, center=(img.shape[1] / 2, img.shape[0] / 2), scale=s)

    # Translation
    T = np.eye(3)
    T[0, 2] = (random.random() * 2 - 1) * translate[0] * img.shape[0] + border  # x translation (pixels)
    T[1, 2] = (random.random() * 2 - 1) * translate[1] * img.shape[1] + border  # y translation (pixels)

    # Shear
    S = np.eye(3)
    S[0, 1] = math.tan((random.random() * (shear[1] - shear[0]) + shear[0]) * math.pi / 180)  # x shear (deg)
    S[1, 0] = math.tan((random.random() * (shear[1] - shear[0]) + shear[0]) * math.pi / 180)  # y shear (deg)

    M = S @ T @ R  # Combined rotation matrix. ORDER IS IMPORTANT HERE!!
    imw = cv2.warpAffine(img, M[:2], dsize=(width, height), flags=cv2.INTER_LINEAR,
                              borderValue=borderValue)  # BGR order borderValue

    # Return warped points also
    if len(targets) > 0:
        n = targets.shape[0]
        points = targets[:, 1:5].copy()
        area0 = (points[:, 2] - points[:, 0]) * (points[:, 3] - points[:, 1])

        # warp points
        xy = np.ones((n * 4, 3))
        xy[:, :2] = points[:, [0, 1, 2, 3, 0, 3, 2, 1]].reshape(n * 4, 2)  # x1y1, x2y2, x1y2, x2y1
        xy = (xy @ M.T)[:, :2].reshape(n, 8)

        # create new boxes
        x = xy[:, [0, 2, 4, 6]]
        y = xy[:, [1, 3, 5, 7]]
        xy = np.concatenate((x.min(1), y.min(1), x.max(1), y.max(1))).reshape(4, n).T

        # # apply angle-based reduction of bounding boxes
        # radians = a * math.pi / 180
        # reduction = max(abs(math.sin(radians)), abs(math.cos(radians))) ** 0.5
        # x = (xy[:, 2] + xy[:, 0]) / 2
        # y = (xy[:, 3] + xy[:, 1]) / 2
        # w = (xy[:, 2] - xy[:, 0]) * reduction
        # h = (xy[:, 3] - xy[:, 1]) * reduction
        # xy = np.concatenate((x - w / 2, y - h / 2, x + w / 2, y + h / 2)).reshape(4, n).T

        # reject warped points outside of image
        xy[:, [0, 2]] = xy[:, [0, 2]].clip(0, width)
        xy[:, [1, 3]] = xy[:, [1, 3]].clip(0, height)
        w = xy[:, 2] - xy[:, 0]
        h = xy[:, 3] - xy[:, 1]
        area = w * h
        ar = np.maximum(w / (h + 1e-16), h / (w + 1e-16))
        i = (w > 4) & (h > 4) & (area / (area0 + 1e-16) > 0.1) & (ar < 10)

        targets = targets[i]
        targets[:, 1:5] = xy[i]

    return imw, targets


def convert_images2bmp():
    # cv2.imread() jpg at 230 img/s, *.bmp at 400 img/s
    for path in ['../coco/images/val2014/', '../coco/images/train2014/']:
        folder = os.sep + Path(path).name
        output = path.replace(folder, folder + 'bmp')
        if os.path.exists(output):
            shutil.rmtree(output)  # delete output folder
        os.makedirs(output)  # make new output folder

        for f in tqdm(glob.glob('%s*.jpg' % path)):
            save_name = f.replace('.jpg', '.bmp').replace(folder, folder + 'bmp')
            cv2.imwrite(save_name, cv2.imread(f))

    for label_path in ['../coco/trainvalno5k.txt', '../coco/5k.txt']:
        with open(label_path, 'r') as file:
            lines = file.read()
        lines = lines.replace('2014/', '2014bmp/').replace('.jpg', '.bmp').replace(
            '/Users/glennjocher/PycharmProjects/', '../')
        with open(label_path.replace('5k', '5k_bmp'), 'w') as file:
            file.write(lines)

增加一段代码，上面已经添加了。
在class LoadWebcam的init中 ，修改为：

  def __init__(self, img_size=416):
        vid_formats = ['.mov', '.avi', '.mp4']#64
        files = []#64

        self.cam = cv2.VideoCapture(0)
        self.height = img_size
        videos = [x for x in files if os.path.splitext(x)[-1].lower() in vid_formats]#64
        self.mode = 'images'#64
        if any(videos):#64
            self.new_video(videos[0])  # new video#64
        else:#64
            self.cap = None#64

报错解决，此时摄像头会把拍到的脸变成图片的形式存在output中

6. 可视化

python -c from utils import utils;utils.plot_results()
比如face和mask
准确度precision=tp/(tp+fp) 识别出来的face占识别出来face和识别出来的面具的比率
召回率recall=tp/(tp+fn)识别出来的face占face总数的比率

完整代码百度云链接

链接：https://pan.baidu.com/s/1J5lEbPULafJNCTFMuLQ88g
提取码：74sb