前言
缺陷檢測(cè)是工業(yè)上非常重要的一個(gè)應(yīng)用,由于缺陷多種多樣,傳統(tǒng)的機(jī)器視覺算法很難做到對(duì)缺陷特征完整的建模和遷移,復(fù)用性不大,要求區(qū)分工況,這會(huì)浪費(fèi)大量的人力成本。深度學(xué)習(xí)在特征提取和定位上取得了非常好的效果,越來越多的學(xué)者和工程人員開始將深度學(xué)習(xí)算法引入到缺陷檢測(cè)領(lǐng)域中。
導(dǎo)師一直鼓勵(lì)小編做一些小項(xiàng)目,將學(xué)習(xí)與動(dòng)手相結(jié)合。于是最近小編找來了某個(gè)大數(shù)據(jù)競賽中的一道缺陷檢測(cè)題目,在開源目標(biāo)檢測(cè)框架的基礎(chǔ)上實(shí)現(xiàn)了一個(gè)用于布匹瑕疵檢測(cè)的模型?,F(xiàn)將過程稍作總結(jié),供各位同學(xué)參考。
問題簡介
01
1
實(shí)際背景
布匹的疵點(diǎn)檢測(cè)是紡織工業(yè)中的一個(gè)十分重要的環(huán)節(jié)。當(dāng)前,在紡織工業(yè)的布匹缺陷檢測(cè)領(lǐng)域,人工檢測(cè)仍然是主要的質(zhì)量檢測(cè)方式。而近年來由于人力成本的提升,以及人工檢測(cè)存在的檢測(cè)速度慢、漏檢率高、一致性差、人員流動(dòng)率高等問題,越來越多的工廠開始利用機(jī)器來代替人工進(jìn)行質(zhì)檢,以提高生產(chǎn)效率,節(jié)省人力成本。
2
題目內(nèi)容
開發(fā)出高效準(zhǔn)確的深度學(xué)習(xí)算法,檢驗(yàn)布匹表面是否存在缺陷,如果存在缺陷,請(qǐng)標(biāo)注出缺陷的類型和位置。
3
數(shù)據(jù)分析
? 題目數(shù)據(jù)集提供了9576張圖片用于訓(xùn)練,其中有瑕疵圖片5913張,無瑕疵圖片3663張。
? 瑕疵共分為15個(gè)類別。分別為:沾污、錯(cuò)花、水卬、花毛、縫頭、縫頭印、蟲粘、破洞、褶子、織疵、漏印、蠟斑、色差、網(wǎng)折、其它
? 尺寸:4096 * 1696
算法分享
02
本文算法基于開源框架YOLOv5,原框架代碼請(qǐng)前往https://github.com/ultralytics/yolov5查看,針對(duì)這次問題做出的修改和調(diào)整部分代碼請(qǐng)繼續(xù)向下閱讀。
1.框架選擇
比較流行的算法可以分為兩類,一類是基于Region Proposal的R-CNN系算法(R-CNN,F(xiàn)ast R-CNN, Faster R-CNN等),它們是two-stage的,需要先算法產(chǎn)生目標(biāo)候選框,也就是目標(biāo)位置,然后再對(duì)候選框做分類與回歸。而另一類是Yolo,SSD這類one-stage算法,其僅僅使用一個(gè)卷積神經(jīng)網(wǎng)絡(luò)CNN直接預(yù)測(cè)不同目標(biāo)的類別與位置。
第一類方法是準(zhǔn)確度高一些,但是速度慢,但是第二類算法是速度快,但是準(zhǔn)確性要低一些。考慮本次任務(wù)時(shí)間限制和小編電腦性能,本次小編采用了單階段YOLOV5的方案。
YOLO直接在輸出層回歸bounding box的位置和bounding box所屬類別,從而實(shí)現(xiàn)one-stage。通過這種方式,Yolo可實(shí)現(xiàn)45幀每秒的運(yùn)算速度,完全能滿足實(shí)時(shí)性要求(達(dá)到24幀每秒,人眼就認(rèn)為是連續(xù)的)。
2.環(huán)境配置(參考自 YOLOv5 requirements)
Cython numpy==1.17 opencv-python torch》=1.4 matplotlib pillow tensorboard PyYAML》=5.3torchvisionscipytqdmgit+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI
3.數(shù)據(jù)預(yù)處理
· 數(shù)據(jù)集文件結(jié)構(gòu)
· 標(biāo)注格式說明
· YOLO要求訓(xùn)練數(shù)據(jù)文件結(jié)構(gòu):
· 比賽數(shù)據(jù)格式 -》 YOLO數(shù)據(jù)格式:
(針對(duì)本問題原創(chuàng)代碼)
for fold in [0]: val_index = index[len(index) * fold // 5:len(index) * (fold + 1) // 5] print(len(val_index)) for num, name in enumerate(name_list): print(c_list[num], x_center_list[num], y_center_list[num], w_list[num], h_list[num]) row = [c_list[num], x_center_list[num], y_center_list[num], w_list[num], h_list[num]] if name in val_index: path2save = ‘val/’ else: path2save = ‘train/’ if not os.path.exists(‘convertor/fold{}/labels/’.format(fold) + path2save): os.makedirs(‘convertor/fold{}/labels/’.format(fold) + path2save) with open(‘convertor/fold{}/labels/’.format(fold) + path2save + name.split(‘。’)[0] + “.txt”, ‘a(chǎn)+’) as f: for data in row: f.write(‘{} ’.format(data)) f.write(‘
’) if not os.path.exists(‘convertor/fold{}/images/{}’.format(fold, path2save)): os.makedirs(‘convertor/fold{}/images/{}’.format(fold, path2save)) sh.copy(os.path.join(image_path, name.split(‘?!?], name), ‘convertor/fold{}/images/{}/{}’.format(fold, path2save, name))
4.超參數(shù)設(shè)置(針對(duì)本問題原創(chuàng)代碼)
# Hyperparameters hyp = {‘lr0’: 0.01, # initial learning rate (SGD=1E-2, Adam=1E-3) ‘momentum’: 0.937, # SGD momentum ‘weight_decay’: 5e-4, # optimizer weight decay ‘giou’: 0.05, # giou loss gain ‘cls’: 0.58, # cls loss gain ‘cls_pw’: 1.0, # cls BCELoss positive_weight ‘obj’: 1.0, # obj loss gain (*=img_size/320 if img_size != 320) ‘obj_pw’: 1.0, # obj BCELoss positive_weight ‘iou_t’: 0.20, # iou training threshold ‘a(chǎn)nchor_t’: 4.0, # anchor-multiple threshold ‘fl_gamma’: 0.0, # focal loss gamma (efficientDet default is gamma=1.5) ‘hsv_h’: 0.014, # image HSV-Hue augmentation (fraction) ‘hsv_s’: 0.68, # image HSV-Saturation augmentation (fraction) ‘hsv_v’: 0.36, # image HSV-Value augmentation (fraction) ‘degrees’: 0.0, # image rotation (+/- deg) ‘translate’: 0.0, # image translation (+/- fraction) ‘scale’: 0.5, # image scale (+/- gain) ‘shear’: 0.0} # image shear (+/- deg)}
5.模型核心代碼(針對(duì)本問題原創(chuàng)代碼)
import argparse from models.experimental import * class Detect(nn.Module): def __init__(self, nc=80, anchors=()): # detection layer super(Detect, self).__init__() self.stride = None # strides computed during build self.nc = nc # number of classes self.no = nc + 5 # number of outputs per anchor self.nl = len(anchors) # number of detection layers self.na = len(anchors[0]) // 2 # number of anchors self.grid = [torch.zeros(1)] * self.nl # init grid a = torch.tensor(anchors).float().view(self.nl, -1, 2) self.register_buffer(‘a(chǎn)nchors’, a) # shape(nl,na,2) self.register_buffer(‘a(chǎn)nchor_grid’, a.clone().view(self.nl, 1, -1, 1, 1, 2)) # shape(nl,1,na,1,1,2) self.export = False # onnx export def forward(self, x): # x = x.copy() # for profiling z = [] # inference output self.training |= self.export for i in range(self.nl): bs, _, ny, nx = x[i].shape # x(bs,255,20,20) to x(bs,3,20,20,85) x[i] = x[i].view(bs, self.na, self.no, ny, nx).permute(0, 1, 3, 4, 2).contiguous() if not self.training: # inference if self.grid[i].shape[2:4] != x[i].shape[2:4]: self.grid[i] = self._make_grid(nx, ny).to(x[i].device) y = x[i].sigmoid() y[。。., 0:2] = (y[。。., 0:2] * 2. - 0.5 + self.grid[i].to(x[i].device)) * self.stride[i] # xy y[。。., 2:4] = (y[。。., 2:4] * 2) ** 2 * self.anchor_grid[i] # wh z.append(y.view(bs, -1, self.no)) return x if self.training else (torch.cat(z, 1), x) @staticmethod def _make_grid(nx=20, ny=20): yv, xv = torch.meshgrid([torch.arange(ny), torch.arange(nx)]) return torch.stack((xv, yv), 2).view((1, 1, ny, nx, 2)).float() class Model(nn.Module): def __init__(self, model_cfg=‘yolov5s.yaml’, ch=3, nc=None): # model, input channels, number of classes super(Model, self).__init__() if type(model_cfg) is dict: self.md = model_cfg # model dict else: # is *.yaml import yaml # for torch hub with open(model_cfg) as f: self.md = yaml.load(f, Loader=yaml.FullLoader) # model dict # Define model if nc and nc != self.md[‘nc’]: print(‘Overriding %s nc=%g with nc=%g’ % (model_cfg, self.md[‘nc’], nc)) self.md[‘nc’] = nc # override yaml value self.model, self.save = parse_model(self.md, ch=[ch]) # model, savelist, ch_out # print([x.shape for x in self.forward(torch.zeros(1, ch, 64, 64))]) # Build strides, anchors m = self.model[-1] # Detect() if isinstance(m, Detect): s = 128 # 2x min stride m.stride = torch.tensor([s / x.shape[-2] for x in self.forward(torch.zeros(1, ch, s, s))]) # forward m.anchors /= m.stride.view(-1, 1, 1) check_anchor_order(m) self.stride = m.stride self._initialize_biases() # only run once # print(‘Strides: %s’ % m.stride.tolist()) # Init weights, biases torch_utils.initialize_weights(self) self._initialize_biases() # only run once torch_utils.model_info(self) print(‘’) def forward(self, x, augment=False, profile=False): if augment: img_size = x.shape[-2:] # height, width s = [0.83, 0.67] # scales #1.2 0.83 y = [] for i, xi in enumerate((x, torch_utils.scale_img(x.flip(3), s[0]), # flip-lr and scale torch_utils.scale_img(x, s[1]), # scale )): # cv2.imwrite(‘img%g.jpg’ % i, 255 * xi[0].numpy().transpose((1, 2, 0))[:, :, ::-1]) y.append(self.forward_once(xi)[0]) y[1][。。., :4] /= s[0] # scale y[1][。。., 0] = img_size[1] - y[1][。。., 0] # flip lr y[2][。。., :4] /= s[1] # scale return torch.cat(y, 1), None # augmented inference, train else: return self.forward_once(x, profile) # single-scale inference, train def forward_once(self, x, profile=False): y, dt = [], [] # outputs for m in self.model: if m.f != -1: # if not from previous layer x = y[m.f] if isinstance(m.f, int) else [x if j == -1 else y[j] for j in m.f] # from earlier layers
if profile: try: import thop o = thop.profile(m, inputs=(x,), verbose=False)[0] / 1E9 * 2 # FLOPS except: o = 0 t = torch_utils.time_synchronized() for _ in range(10): _ = m(x) dt.append((torch_utils.time_synchronized() - t) * 100) print(‘%10.1f%10.0f%10.1fms %-40s’ % (o, m.np, dt[-1], m.type)) x = m(x) # run y.append(x if m.i in self.save else None) # save output
if profile: print(‘%.1fms total’ % sum(dt)) return x
def _initialize_biases(self, cf=None): # initialize biases into Detect(), cf is class frequency # cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1. m = self.model[-1] # Detect() module for f, s in zip(m.f, m.stride): # from mi = self.model[f % m.i] b = mi.bias.view(m.na, -1) # conv.bias(255) to (3,85) b[:, 4] += math.log(8 / (640 / s) ** 2) # obj (8 objects per 640 image) b[:, 5:] += math.log(0.6 / (m.nc - 0.99)) if cf is None else torch.log(cf / cf.sum()) # cls mi.bias = torch.nn.Parameter(b.view(-1), requires_grad=True)
def _print_biases(self): m = self.model[-1] # Detect() module for f in sorted([x % m.i for x in m.f]): # from b = self.model[f].bias.detach().view(m.na, -1).T # conv.bias(255) to (3,85) print((‘%g Conv2d.bias:’ + ‘%10.3g’ * 6) % (f, *b[:5].mean(1).tolist(), b[5:].mean()))
# def _print_weights(self): # for m in self.model.modules(): # if type(m) is Bottleneck: # print(‘%10.3g’ % (m.w.detach().sigmoid() * 2)) # shortcut weights
def fuse(self): # fuse model Conv2d() + BatchNorm2d() layers print(‘Fusing layers.。。 ’, end=‘’) for m in self.model.modules(): if type(m) is Conv: m.conv = torch_utils.fuse_conv_and_bn(m.conv, m.bn) # update conv m.bn = None # remove batchnorm m.forward = m.fuseforward # update forward torch_utils.model_info(self) return self
def parse_model(md, ch): # model_dict, input_channels(3) print(‘
%3s%18s%3s%10s %-40s%-30s’ % (‘’, ‘from’, ‘n’, ‘params’, ‘module’, ‘a(chǎn)rguments’)) anchors, nc, gd, gw = md[‘a(chǎn)nchors’], md[‘nc’], md[‘depth_multiple’], md[‘width_multiple’] na = (len(anchors[0]) // 2) # number of anchors no = na * (nc + 5) # number of outputs = anchors * (classes + 5)
layers, save, c2 = [], [], ch[-1] # layers, savelist, ch out for i, (f, n, m, args) in enumerate(md[‘backbone’] + md[‘head’]): # from, number, module, args m = eval(m) if isinstance(m, str) else m # eval strings for j, a in enumerate(args): try: args[j] = eval(a) if isinstance(a, str) else a # eval strings except: pass
n = max(round(n * gd), 1) if n 》 1 else n # depth gain if m in [nn.Conv2d, Conv, PW_Conv,Bottleneck, SPP, DWConv, MixConv2d, Focus, CrossConv, BottleneckCSP, C3, BottleneckMOB]: c1, c2 = ch[f], args[0]
# Normal # if i 》 0 and args[0] != no: # channel expansion factor # ex = 1.75 # exponential (default 2.0) # e = math.log(c2 / ch[1]) / math.log(2) # c2 = int(ch[1] * ex ** e) # if m != Focus: c2 = make_divisible(c2 * gw, 8) if c2 != no else c2
# Experimental # if i 》 0 and args[0] != no: # channel expansion factor # ex = 1 + gw # exponential (default 2.0) # ch1 = 32 # ch[1] # e = math.log(c2 / ch1) / math.log(2) # level 1-n # c2 = int(ch1 * ex ** e) # if m != Focus: # c2 = make_divisible(c2, 8) if c2 != no else c2
args = [c1, c2, *args[1:]] if m in [BottleneckCSP, C3]: args.insert(2, n) n = 1 elif m is nn.BatchNorm2d: args = [ch[f]] elif m is Concat: c2 = sum([ch[-1 if x == -1 else x + 1] for x in f]) elif m is Detect: f = f or list(reversed([(-1 if j == i else j - 1) for j, x in enumerate(ch) if x == no])) else: c2 = ch[f]
m_ = nn.Sequential(*[m(*args) for _ in range(n)]) if n 》 1 else m(*args) # module t = str(m)[8:-2].replace(‘__main__.’, ‘’) # module type np = sum([x.numel() for x in m_.parameters()]) # number params m_.i, m_.f, m_.type, m_.np = i, f, t, np # attach index, ‘from’ index, type, number params print(‘%3s%18s%3s%10.0f %-40s%-30s’ % (i, f, n, np, t, args)) # print save.extend(x % i for x in ([f] if isinstance(f, int) else f) if x != -1) # append to savelist layers.append(m_) ch.append(c2) return nn.Sequential(*layers), sorted(save)
if __name__ == ‘__main__’: parser = argparse.ArgumentParser() parser.add_argument(‘--cfg’, type=str, default=‘yolov5s.yaml’, help=‘model.yaml’) parser.add_argument(‘--device’, default=‘’, help=‘cuda device, i.e. 0 or 0,1,2,3 or cpu’) opt = parser.parse_args() opt.cfg = check_file(opt.cfg) # check file device = torch_utils.select_device(opt.device)
# Create model model = Model(opt.cfg).to(device) model.train()
訓(xùn)練截圖
6.測(cè)試模型并生成結(jié)果(針對(duì)本問題原創(chuàng)代碼)
for *xyxy, conf, cls in det: if save_txt: # Write to file xywh = (xyxy2xywh(torch.tensor(xyxy).view(1, 4)) / gn).view(-1).tolist() # normalized xywh with open(txt_path + ‘.txt’, ‘a(chǎn)’) as f: f.write((‘%g ’ * 5 + ‘
’) % (cls, *xywh)) # label format # write to json if save_json: name = os.path.split(txt_path)[-1] print(name)
x1, y1, x2, y2 = float(xyxy[0]), float(xyxy[1]), float( xyxy[2]), float(xyxy[3]) bbox = [x1, y1, x2, y2] img_name = name conf = float(conf)
#add solution remove other result.append({ ‘name’: img_name + ‘.jpg’, ‘category’: int(cls + 1), ‘bbox’: bbox, ‘score’: conf })
7.結(jié)果展示
后記
03
針對(duì)布匹瑕疵檢測(cè)問題,我們首先分析了題目要求,確定了我們的任務(wù)是檢測(cè)到布匹中可能存在的瑕疵,對(duì)其進(jìn)行分類并將其在圖片中標(biāo)注出來。接下來針對(duì)問題要求我們選擇了合適的目標(biāo)檢測(cè)框架YOLOv5,并按照YOLOv5的格式要求對(duì)數(shù)據(jù)集和標(biāo)注進(jìn)行了轉(zhuǎn)換。然后我們根據(jù)問題規(guī)模設(shè)置了合適的超參數(shù),采用遷移學(xué)習(xí)的思想,基于官方的預(yù)訓(xùn)練模型進(jìn)行訓(xùn)練以加快收斂速度。模型訓(xùn)練好以后,即可在驗(yàn)證集上驗(yàn)證我們模型的性能和準(zhǔn)確性。
回顧整個(gè)過程我們可以發(fā)現(xiàn),在越來越多的優(yōu)秀目標(biāo)檢測(cè)框架被提出并開源之后,目標(biāo)檢測(cè)模型的實(shí)現(xiàn)門檻越來越低,我們可以很輕松的借用這些框架搭建模型來解決現(xiàn)實(shí)生活中的缺陷檢測(cè)問題,深度學(xué)習(xí)的應(yīng)用并沒有我們想象的那么復(fù)雜。
當(dāng)然,若想得到針對(duì)某個(gè)具體問題表現(xiàn)更加優(yōu)秀的模型,還需要我們根據(jù)具體問題的具體特點(diǎn)對(duì)模型進(jìn)行修正調(diào)優(yōu)。例如針對(duì)本次布匹缺陷檢測(cè)數(shù)據(jù)集中部分缺陷種類樣本數(shù)量少、缺陷目標(biāo)較小的問題,我們可以通過過采樣種類較少的樣本、數(shù)據(jù)增廣、增加anchor的數(shù)量等方法來進(jìn)一步提高模型的準(zhǔn)確率。如果有同學(xué)對(duì)該問題感興趣,想要進(jìn)一步了解或在代碼理解、環(huán)境配置等各方面存在疑問的話,歡迎通過文末郵箱聯(lián)系小編,小編在這里期待與您交流。
文案:張宇(華中科技大學(xué)管理學(xué)院本科二年級(jí))指導(dǎo)老師:曹菁菁(武漢理工大學(xué)物流工程學(xué)院)排版:程欣悅(荊楚理工學(xué)院本科三年級(jí))審稿:張宇(華中科技大學(xué)管理學(xué)院本科二年級(jí))。
—版權(quán)聲明—
來源:數(shù)據(jù)魔術(shù)師 作者:張宇
僅用于學(xué)術(shù)分享,版權(quán)屬于原作者。
若有侵權(quán),請(qǐng)聯(lián)系微信號(hào):刪除或修改!
編輯:jq
-
機(jī)器視覺
+關(guān)注
關(guān)注
163文章
4596瀏覽量
122890 -
數(shù)據(jù)集
+關(guān)注
關(guān)注
4文章
1224瀏覽量
25446 -
深度學(xué)習(xí)
+關(guān)注
關(guān)注
73文章
5561瀏覽量
122794
原文標(biāo)題:深度學(xué)習(xí)實(shí)戰(zhàn)之布匹缺陷檢測(cè)
文章出處:【微信號(hào):vision263com,微信公眾號(hào):新機(jī)器視覺】歡迎添加關(guān)注!文章轉(zhuǎn)載請(qǐng)注明出處。
發(fā)布評(píng)論請(qǐng)先 登錄
在薄膜框架上提供的 PIN 二極管芯片 skyworksinc

基于FPGA的SSD目標(biāo)檢測(cè)算法設(shè)計(jì)

請(qǐng)問嘉楠平臺(tái)k230訓(xùn)練的跌倒檢測(cè)模型里面的設(shè)置是怎么樣的?
基于LockAI視覺識(shí)別模塊:C++目標(biāo)檢測(cè)
RV1126 實(shí)現(xiàn)人臉檢測(cè)方案

AI模型部署邊緣設(shè)備的奇妙之旅:目標(biāo)檢測(cè)模型
案例分享 ▏基于HZ-EVM-RK3576開發(fā)板實(shí)現(xiàn)YOLOv5目標(biāo)檢測(cè)應(yīng)用

大語言模型開發(fā)框架是什么
X-RAY檢測(cè)設(shè)備用于檢測(cè)集成電路缺陷瑕疵

YOLOv10自定義目標(biāo)檢測(cè)之理論+實(shí)踐

手把手教你如何自制目標(biāo)檢測(cè)框架

在樹莓派上部署YOLOv5進(jìn)行動(dòng)物目標(biāo)檢測(cè)的完整流程

在目標(biāo)檢測(cè)中大物體的重要性

工業(yè)主板在服裝紡織瑕疵檢測(cè)中的應(yīng)用

評(píng)論