@Author：Runsen

YOLO 是目前最火爆的的計(jì)算機(jī)視覺(jué)算法之一，今天使用的數(shù)據(jù)集來(lái)源：https://www.kaggle.com/andrewmvd/face-mask-detection

這是數(shù)據(jù)集可以創(chuàng)建一個(gè)模型來(lái)檢測(cè)戴口罩、不戴口罩或不正確戴口罩的人。

該數(shù)據(jù)集包含屬于3 個(gè)類別的853張圖像，以及它們的 PASCAL VOC 格式的邊界框。分別是

• 帶口罩；
• 不戴口罩；
• 口罩佩戴不正確。

首先將voc形式變成yolo形式，voc是xml文件，需要通過(guò)ElementTree進(jìn)行解析，得到bnd_box ，具體解析代碼如下

import os from xml.etree.ElementTree import ElementTreedef xml_to_yolo(path):root = ElementTree().parse(path)img_path = root.find('filename').text.replace('png','txt')img_size = root.find('size')width = int(img_size.find('width').text)height = int(img_size.find('height').text)with open('../dataset/label/' + img_path, 'w') as f:lines = []for node in root.findall('object'):object_ = dict(class_=None, x=None, y=None, width=None, height=None)# classclass_name = node.find('name').textif(class_name == 'without_mask'):object_['class_'] = '0'elif(class_name == 'with_mask'):object_['class_'] = '1'else:object_['class_'] = '2'# bounding boxbnd_box = node.find("bndbox")x_min = float(bnd_box[0].text)y_min = float(bnd_box[1].text)x_max = float(bnd_box[2].text)y_max = float(bnd_box[3].text)dw = float(1/width)dh = float(1/height)w = float(x_max - x_min)h = float(y_max - y_min)x = float((x_min + x_max)/2 -1)y = float((y_min + y_max)/2 -1)w = float(w * dw)h = float(h * dh)x = float(x * dw)y = float(y * dh)object_['x'] = str(x)object_['y'] = str(y)object_['width'] = str(w)object_['height'] = str(h)line = object_['class_'] + ' ' + object_['x'] + ' ' + object_['y'] + ' ' + object_['width'] + ' ' + object_['height']lines.append(line)lines.append('\n')for line in lines[:-1]:f.write(line) f.closedef process_data():img_paths = []for dirname, _, filenames in os.walk('../dataset/images'):for filename in filenames:img_paths.append(os.path.join('../dataset/images', filename)) # local machine# xml for dirname, _, filenames in os.walk('../dataset/xml'):for filename in filenames:annotation_path = (os.path.join(dirname, filename))xml_to_yolo(annotation_path)process_data()

克隆Github的YOLOv5的代碼：https://github.com/ultralytics/yolov5。

為了跟數(shù)據(jù)集的目錄保持一致

在dataset新建images和labels文件夾存放對(duì)于的文件。

data文件夾下的data.yaml文件

train: ./dataset/images/train val: ./dataset/images/val# number of classes nc: 3# class names names: ['without_mask','with_mask','mask_weared_incorrect']

model文件夾下的yolov5s.yaml文件

# Parameters nc: 3 # number of classes depth_multiple: 0.33 # model depth multiple width_multiple: 0.50 # layer channel multiple anchors:- [10,13, 16,30, 33,23] # P3/8- [30,61, 62,45, 59,119] # P4/16- [116,90, 156,198, 373,326] # P5/32# YOLOv5 backbone backbone:# [from, number, module, args][[-1, 1, Focus, [64, 3]], # 0-P1/2[-1, 1, Conv, [128, 3, 2]], # 1-P2/4[-1, 3, C3, [128]],[-1, 1, Conv, [256, 3, 2]], # 3-P3/8[-1, 9, C3, [256]],[-1, 1, Conv, [512, 3, 2]], # 5-P4/16[-1, 9, C3, [512]],[-1, 1, Conv, [1024, 3, 2]], # 7-P5/32[-1, 1, SPP, [1024, [5, 9, 13]]],[-1, 3, C3, [1024, False]], # 9]# YOLOv5 head head:[[-1, 1, Conv, [512, 1, 1]],[-1, 1, nn.Upsample, [None, 2, 'nearest']],[[-1, 6], 1, Concat, [1]], # cat backbone P4[-1, 3, C3, [512, False]], # 13[-1, 1, Conv, [256, 1, 1]],[-1, 1, nn.Upsample, [None, 2, 'nearest']],[[-1, 4], 1, Concat, [1]], # cat backbone P3[-1, 3, C3, [256, False]], # 17 (P3/8-small)[-1, 1, Conv, [256, 3, 2]],[[-1, 14], 1, Concat, [1]], # cat head P4[-1, 3, C3, [512, False]], # 20 (P4/16-medium)[-1, 1, Conv, [512, 3, 2]],[[-1, 10], 1, Concat, [1]], # cat head P5[-1, 3, C3, [1024, False]], # 23 (P5/32-large)[[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5)]

下載yolov5s.pt對(duì)應(yīng)的權(quán)重，大約14M

訓(xùn)練代碼：python train.py --batch 16 --epochs 50 --data /data/data.yaml --cfg /models/yolov5s.yaml

訓(xùn)練結(jié)果

關(guān)于P R mAP50 和mAP95的指標(biāo)

mAP50，mAP60，mAP70……等等指的是取detector的IoU閾值大于0.5，大于0.6，大于0.7…的比例。數(shù)值越高，越好，如果閾值越大，精度會(huì)變低。

因?yàn)镮oU大于0.95的小于IoU大于0.5

yolov5訓(xùn)練完，會(huì)生成runs的文件夾

訓(xùn)練結(jié)束會(huì)生成train文件夾，第一次訓(xùn)練完成一般默認(rèn)是exp，如果再訓(xùn)練一次，就是exp2。exp19是我在不斷的改batch 和epochs。

50次訓(xùn)練的指標(biāo)結(jié)果

epoch, train/box_loss, train/obj_loss, train/cls_loss, metrics/precision, metrics/recall, metrics/mAP_0.5,metrics/mAP_0.5:0.95, val/box_loss, val/obj_loss, val/cls_loss, x/lr0, x/lr1, x/lr20, 0.11183, 0.061692, 0.036166, 0.0040797, 0.023207, 0.00094919, 0.00015348, 0.088802, 0.044058, 0.029341, 0.00053, 0.00053, 0.095231, 0.090971, 0.06624, 0.024651, 0.44734, 0.13502, 0.11474, 0.0279, 0.067798, 0.037852, 0.022663, 0.0010692, 0.0010692, 0.0903692, 0.070342, 0.052357, 0.02153, 0.85191, 0.20848, 0.22242, 0.077028, 0.055487, 0.027934, 0.019595, 0.0016049, 0.0016049, 0.0855053, 0.0699, 0.043899, 0.020244, 0.87295, 0.16442, 0.20657, 0.078501, 0.055767, 0.024396, 0.018845, 0.0021348, 0.0021348, 0.0806354, 0.064317, 0.04001, 0.018747, 0.83973, 0.19818, 0.20992, 0.068715, 0.055507, 0.022381, 0.018095, 0.0026562, 0.0026562, 0.0757565, 0.060517, 0.040462, 0.01871, 0.92677, 0.26977, 0.31691, 0.13554, 0.046723, 0.020782, 0.017638, 0.0031668, 0.0031668, 0.0708676, 0.055565, 0.037719, 0.017239, 0.56328, 0.42257, 0.34731, 0.14537, 0.042455, 0.020308, 0.015934, 0.0036641, 0.0036641, 0.0659647, 0.051137, 0.036675, 0.01443, 0.70063, 0.59002, 0.56658, 0.28446, 0.037626, 0.018761, 0.012734, 0.0041459, 0.0041459, 0.0610468, 0.051564, 0.036086, 0.010754, 0.75331, 0.59469, 0.57015, 0.21859, 0.044253, 0.018153, 0.010504, 0.00461, 0.00461, 0.056119, 0.054336, 0.036489, 0.01014, 0.81292, 0.53958, 0.57799, 0.28552, 0.037579, 0.018368, 0.009468, 0.0050544, 0.0050544, 0.05115410, 0.052323, 0.034572, 0.0095445, 0.85048, 0.62312, 0.6267, 0.338, 0.031516, 0.017655, 0.010057, 0.005477, 0.005477, 0.04617711, 0.050382, 0.035035, 0.0086806, 0.86829, 0.59491, 0.61092, 0.3379, 0.032883, 0.017608, 0.0090742, 0.0058761, 0.0058761, 0.04117612, 0.046103, 0.035051, 0.0082707, 0.8746, 0.60331, 0.72019, 0.42562, 0.030186, 0.016849, 0.0087845, 0.00625, 0.00625, 0.0361513, 0.045144, 0.032905, 0.0071727, 0.50942, 0.80319, 0.76972, 0.38975, 0.033539, 0.016185, 0.0083475, 0.0065973, 0.0065973, 0.03109714, 0.043005, 0.031431, 0.0076075, 0.94694, 0.59886, 0.75893, 0.45433, 0.028811, 0.016204, 0.0087185, 0.0069167, 0.0069167, 0.02601715, 0.042583, 0.032493, 0.0075345, 0.54628, 0.78986, 0.70483, 0.39052, 0.034077, 0.016426, 0.010784, 0.007207, 0.007207, 0.02090716, 0.044892, 0.032174, 0.0079312, 0.93707, 0.62972, 0.75324, 0.417, 0.031565, 0.015583, 0.0074945, 0.0074674, 0.0074674, 0.01576717, 0.04351, 0.033322, 0.0071257, 0.9655, 0.60346, 0.75329, 0.43919, 0.029984, 0.015935, 0.0071412, 0.0076971, 0.0076971, 0.01059718, 0.041603, 0.032276, 0.0070445, 0.56572, 0.84931, 0.80283, 0.50439, 0.027445, 0.015567, 0.0067956, 0.0077031, 0.0077031, 0.007703119, 0.03878, 0.031363, 0.0064713, 0.57055, 0.85292, 0.81709, 0.54818, 0.02845, 0.014806, 0.006212, 0.0077031, 0.0077031, 0.007703120, 0.037742, 0.030032, 0.0062044, 0.57265, 0.90265, 0.83532, 0.52484, 0.024745, 0.01497, 0.0060971, 0.0074725, 0.0074725, 0.007472521, 0.035583, 0.029891, 0.0062017, 0.6255, 0.95098, 0.87989, 0.55897, 0.026561, 0.014704, 0.0072068, 0.0072361, 0.0072361, 0.007236122, 0.034756, 0.030238, 0.0059138, 0.76515, 0.85503, 0.93007, 0.62612, 0.02538, 0.014571, 0.0066402, 0.0069948, 0.0069948, 0.006994823, 0.033698, 0.030118, 0.0058317, 0.71699, 0.99118, 0.96047, 0.68458, 0.024377, 0.01393, 0.0055267, 0.0067495, 0.0067495, 0.006749524, 0.031946, 0.029104, 0.0052597, 0.80874, 0.98185, 0.97812, 0.6923, 0.024714, 0.014598, 0.0056557, 0.0065013, 0.0065013, 0.006501325, 0.031391, 0.029652, 0.0048984, 0.74641, 0.99789, 0.97365, 0.64525, 0.024137, 0.013945, 0.0055031, 0.0062512, 0.0062512, 0.006251226, 0.029442, 0.028553, 0.005284, 0.77383, 0.94401, 0.96556, 0.72068, 0.022413, 0.013277, 0.0049306, 0.006, 0.006, 0.00627, 0.029758, 0.027931, 0.0052183, 0.85002, 0.95022, 0.98664, 0.72732, 0.022897, 0.013339, 0.005281, 0.0057488, 0.0057488, 0.005748828, 0.027972, 0.028458, 0.0046879, 0.89455, 0.99578, 0.99146, 0.70404, 0.022004, 0.013255, 0.0037246, 0.0054987, 0.0054987, 0.005498729, 0.027527, 0.026557, 0.0043153, 0.95, 0.97609, 0.98743, 0.67707, 0.022925, 0.013022, 0.0033831, 0.0052505, 0.0052505, 0.005250530, 0.028249, 0.028263, 0.0041167, 0.97478, 0.98066, 0.99444, 0.72056, 0.020498, 0.012977, 0.0044037, 0.0050052, 0.0050052, 0.005005231, 0.027591, 0.026807, 0.0039948, 0.99205, 0.97724, 0.99428, 0.68932, 0.021719, 0.01319, 0.0033467, 0.0047639, 0.0047639, 0.004763932, 0.027873, 0.027414, 0.0041063, 0.93928, 0.97948, 0.99237, 0.71491, 0.021192, 0.012699, 0.0026829, 0.0045275, 0.0045275, 0.004527533, 0.026139, 0.025726, 0.0036621, 0.98341, 0.94802, 0.99364, 0.7167, 0.020188, 0.012794, 0.0028303, 0.0042969, 0.0042969, 0.004296934, 0.025655, 0.026828, 0.0033113, 0.97988, 0.9678, 0.99428, 0.7098, 0.019681, 0.012586, 0.0024685, 0.004073, 0.004073, 0.00407335, 0.025622, 0.026283, 0.0033939, 0.98079, 0.96134, 0.9945, 0.73714, 0.020264, 0.012446, 0.0022277, 0.0038567, 0.0038567, 0.003856736, 0.02521, 0.025613, 0.003236, 0.98049, 0.98546, 0.99512, 0.71957, 0.019993, 0.012117, 0.0020501, 0.0036489, 0.0036489, 0.003648937, 0.024812, 0.026664, 0.0031498, 0.98609, 0.98551, 0.99503, 0.73326, 0.018914, 0.012313, 0.0019162, 0.0034503, 0.0034503, 0.003450338, 0.024114, 0.025925, 0.0029253, 0.987, 0.98471, 0.99513, 0.74502, 0.018339, 0.012051, 0.0018452, 0.0032618, 0.0032618, 0.003261839, 0.02411, 0.023745, 0.0030621, 0.98562, 0.981, 0.99503, 0.75325, 0.018894, 0.012225, 0.0019048, 0.0030841, 0.0030841, 0.003084140, 0.023838, 0.02533, 0.002918, 0.98616, 0.98312, 0.99524, 0.7713, 0.018355, 0.011694, 0.0017296, 0.0029179, 0.0029179, 0.002917941, 0.024228, 0.025735, 0.0031651, 0.97865, 0.98523, 0.99515, 0.77026, 0.018076, 0.011688, 0.0021994, 0.0027639, 0.0027639, 0.002763942, 0.023581, 0.025971, 0.0027009, 0.98297, 0.9877, 0.99519, 0.7426, 0.018612, 0.011757, 0.0014865, 0.0026227, 0.0026227, 0.002622743, 0.02345, 0.025302, 0.0028664, 0.97805, 0.98938, 0.99525, 0.76359, 0.018273, 0.011761, 0.0013719, 0.0024948, 0.0024948, 0.002494844, 0.023614, 0.024621, 0.0028453, 0.99051, 0.98734, 0.99537, 0.75192, 0.017944, 0.01167, 0.0012174, 0.0023807, 0.0023807, 0.002380745, 0.023214, 0.024148, 0.0023234, 0.98904, 0.98296, 0.99532, 0.75336, 0.018311, 0.011531, 0.0011202, 0.0022809, 0.0022809, 0.002280946, 0.022992, 0.025107, 0.0026533, 0.98561, 0.98809, 0.99534, 0.76908, 0.017906, 0.011331, 0.0011833, 0.0021958, 0.0021958, 0.002195847, 0.023114, 0.023664, 0.0026155, 0.9892, 0.98336, 0.99531, 0.76103, 0.018336, 0.011521, 0.0011389, 0.0021257, 0.0021257, 0.002125748, 0.02274, 0.02358, 0.002211, 0.98464, 0.98066, 0.99535, 0.76574, 0.017788, 0.011413, 0.0011274, 0.0020709, 0.0020709, 0.002070949, 0.023034, 0.024154, 0.0021493, 0.98092, 0.98945, 0.99529, 0.75593, 0.017796, 0.011467, 0.00096875, 0.0020315, 0.0020315, 0.0020315

測(cè)試代碼：python detect.py --weights runs/train/exp19/weights/best.pt --img 512 --conf 0.4 --source dataset/images/val --save-txt

預(yù)測(cè)結(jié)束，會(huì)在runs文件夾生成detect文件夾

如果加了–save-txt，就有對(duì)應(yīng)的訓(xùn)練樣本的yolo格式

下面的幾個(gè)測(cè)試結(jié)果

附上數(shù)據(jù)集和訓(xùn)練結(jié)果：

鏈接：https://pan.baidu.com/s/1XQtLBo5KPzZYCH2Yqm0S1g
提取碼：toxf

總結(jié)

以上是生活随笔為你收集整理的深度学习和目标检测系列教程 17-300： 3 个类别面罩检测类别数据集训练yolov5s模型的全部?jī)?nèi)容，希望文章能夠幫你解決所遇到的問(wèn)題。

如果覺(jué)得生活随笔網(wǎng)站內(nèi)容還不錯(cuò)，歡迎將生活随笔推薦給好友。