【项目三、车牌检测+识别项目】四、使用LPRNet进行车牌识别

前言

马上要找工作了，想总结下自己做过的几个小项目。

之前已经总结过了我做的第一个项目：xxx病虫害检测项目，github源码地址：HuKai97/FFSSD-ResNet。CSDN讲解地址：

1. 【项目一、xxx病虫害检测项目】1、SSD原理和源码分析
2. 【项目一、xxx病虫害检测项目】2、网络结构尝试改进：Resnet50、SE、CBAM、Feature Fusion
3. 【项目一、xxx病虫害检测项目】3、损失函数尝试：Focal loss

第二个项目：蜂巢检测项目，github源码地址：https://github.com/HuKai97/YOLOv5-ShuffleNetv2。CSDN讲解地址：

【项目二、蜂巢检测项目】一、串讲各类经典的卷积网络：InceptionV1-V4、ResNetV1-V2、MobileNetV1-V3、ShuffleNetV1-V2、ResNeXt、Xception。
【项目二、蜂巢检测项目】二、模型改进：YOLOv5s-ShuffleNetV2。

如果对YOLOv5不熟悉的同学可以先看看我写的YOLOv5源码讲解CSDN:【YOLOV5-5.x 源码讲解】整体项目文件导航，注释版YOLOv5源码我也开源在了Github上：HuKai97/yolov5-5.x-annotations，欢迎大家star!

之前一直在学习OCR相关的东西，就想着能不能做一个车牌识别的项目出来，刚好车牌检测也好做，直接用v5就可以了。我的打算是做一个轻量级的车牌识别项目，检测网络用的是YOLOv5s，识别网络有的是LPRNet。

这一节主要介绍下怎么训练LPRNet车牌识别模型。

车牌识别项目所有讲解：

1. 【项目三、车牌检测+识别项目】一、CCPD车牌数据集转为YOLOv5格式和LPRNet格式
2. 【项目三、车牌检测+识别项目】二、使用YOLOV5进行车牌检测
3. 【项目三、车牌检测+识别项目】三、LPRNet车牌识别网络原理和核心源码解读
4. 【项目三、车牌检测+识别项目】四、使用LPRNet进行车牌识别

代码已全部上传GitHub：https://github.com/HuKai97/YOLOv5-LPRNet-Licence-Recognition

一、数据集

在项目文件 YOLOv5-LRPNet-Licence-Recognition 的同级目录建一个datasets文件，数据集这里写的很清楚：【项目三、车牌检测+识别项目】一、CCPD车牌数据集转为YOLOv5格式和LPRNet格式，制作好放在在datasets里面，我一般不会把数据和代码放在一起，那样pycharm打开非常慢。

二、训练

在tools下的train_lprnet.py，主要配置：

def get_parser():
    parser = argparse.ArgumentParser(description='parameters to train net')
    parser.add_argument('--max_epoch', default=100, help='epoch to train the network')
    parser.add_argument('--img_size', default=[94, 24], help='the image size')
    parser.add_argument('--train_img_dirs', default=r"", help='the train images path')
    parser.add_argument('--test_img_dirs', default=r"", help='the test images path')
    parser.add_argument('--dropout_rate', default=0.5, help='dropout rate.')
    parser.add_argument('--learning_rate', default=0.01, help='base value of learning rate.')
    parser.add_argument('--lpr_max_len', default=8, help='license plate number max length.')
    parser.add_argument('--train_batch_size', default=128, help='training batch size.')
    parser.add_argument('--test_batch_size', default=128, help='testing batch size.')
    parser.add_argument('--phase_train', default=True, type=bool, help='train or test phase flag.')
    parser.add_argument('--num_workers', default=8, type=int, help='Number of workers used in dataloading')
    parser.add_argument('--cuda', default=True, type=bool, help='Use cuda to train model')
    parser.add_argument('--resume_epoch', default=0, type=int, help='resume iter for retraining')
    parser.add_argument('--save_interval', default=500, type=int, help='interval for save model state dict')
    parser.add_argument('--test_interval', default=500, type=int, help='interval for evaluate')
    parser.add_argument('--momentum', default=0.9, type=float, help='momentum')
    parser.add_argument('--weight_decay', default=2e-5, type=float, help='Weight decay for SGD')
    parser.add_argument('--lr_schedule', default=[20, 40, 60, 80, 100], help='schedule for learning rate.')
    parser.add_argument('--save_folder', default=r'',
                        help='Location to save checkpoint models')
    parser.add_argument('--pretrained_model', default='', help='no pretrain')

另外，我是没加载预训练权重的，一开始训不好，后来发现调好learning_rate和lr_schedule这两个参数，一般都能训到不错的分数（93+）。

三、验证

在tools下的test_lprnet.py，主要配置：

def get_parser():
    parser = argparse.ArgumentParser(description='parameters to train net')
    parser.add_argument('--img_size', default=[94, 24], help='the image size')
    parser.add_argument('--test_img_dirs', default=r"", help='the test images path')
    parser.add_argument('--dropout_rate', default=0, help='dropout rate.')
    parser.add_argument('--lpr_max_len', default=8, help='license plate number max length.')
    parser.add_argument('--test_batch_size', default=100, help='testing batch size.')
    parser.add_argument('--phase_train', default=False, type=bool, help='train or test phase flag.')
    parser.add_argument('--num_workers', default=0, type=int, help='Number of workers used in dataloading')
    parser.add_argument('--cuda', default=True, type=bool, help='Use cuda to train model')
    parser.add_argument('--show', default=False, type=bool, help='show test image and its predict result or not.')
    parser.add_argument('--pretrained_model', default=r'', help='pretrained base model')

LPRNet算法性能：

model	数据集	epochs	acc	size
LPRNet	val	100	94.33	1.7M
LPRNet	test	100	94.30	1.7M

整个模型（YOLOv5+LPRNet）速度：47.6FPS（970 GPU）

四、测试结果

当然也有检测的不好的：

五、推理代码

main.py

import argparse

import torch.backends.cudnn as cudnn

from models.experimental import *
from utils.datasets import *
from utils.utils import *
from models.LPRNet import *

def detect(save_img=False):
    classify, out, source, det_weights, rec_weights, view_img, save_txt, imgsz = \
        opt.classify, opt.output, opt.source, opt.det_weights, opt.rec_weights,  opt.view_img, opt.save_txt, opt.img_size
    webcam = source == '0' or source.startswith('rtsp') or source.startswith('http') or source.endswith('.txt')

    # Initialize
    device = torch_utils.select_device(opt.device)
    if os.path.exists(out):
        shutil.rmtree(out)  # delete rec_result folder
    os.makedirs(out)  # make new rec_result folder
    half = device.type != 'cpu'  # half precision only supported on CUDA

    # Load yolov5 model
    model = attempt_load(det_weights, map_location=device)  # load FP32 model
    print("load det pretrained model successful!")
    imgsz = check_img_size(imgsz, s=model.stride.max())  # check img_size
    if half:
        model.half()  # to FP16

    # Second-stage classifier  也就是rec 字符识别
    if classify:
        modelc = LPRNet(lpr_max_len=8, phase=False, class_num=len(CHARS), dropout_rate=0).to(device)
        modelc.load_state_dict(torch.load(rec_weights, map_location=torch.device('cpu')))
        print("load rec pretrained model successful!")
        modelc.to(device).eval()

    # Set Dataloader
    vid_path, vid_writer = None, None
    if webcam:
        view_img = True
        cudnn.benchmark = True  # set True to speed up constant image size demo
        dataset = LoadStreams(source, img_size=imgsz)
    else:
        save_img = True
        dataset = LoadImages(source, img_size=imgsz)

    # Get names and colors
    names = model.module.names if hasattr(model, 'module') else model.names
    colors = [[random.randint(0, 255) for _ in range(3)] for _ in range(len(names))]

    # Run demo
    t0 = time.time()
    img = torch.zeros((1, 3, imgsz, imgsz), device=device)  # init img
    _ = model(img.half() if half else img) if device.type != 'cpu' else None  # run once
    for path, img, im0s, vid_cap in dataset:
        img = torch.from_numpy(img).to(device)
        img = img.half() if half else img.float()  # uint8 to fp16/32
        img /= 255.0  # 0 - 255 to 0.0 - 1.0
        if img.ndimension() == 3:
            img = img.unsqueeze(0)

        # Inference
        t1 = torch_utils.time_synchronized()
        pred = model(img, augment=opt.augment)[0]
        # Apply NMS
        pred = non_max_suppression(pred, opt.conf_thres, opt.iou_thres, classes=opt.classes, agnostic=opt.agnostic_nms)
        t2 = torch_utils.time_synchronized()

        # Apply Classifier
        if classify:
            pred, plat_num = apply_classifier(pred, modelc, img, im0s)

        # Process detections
        for i, det in enumerate(pred):  # detections per image
            if webcam:  # batch_size >= 1
                p, s, im0 = path[i], '%g: ' % i, im0s[i].copy()
            else:
                p, s, im0 = path, '', im0s

            save_path = str(Path(out) / Path(p).name)
            txt_path = str(Path(out) / Path(p).stem) + ('_%g' % dataset.frame if dataset.mode == 'video' else '')
            s += '%gx%g ' % img.shape[2:]  # print string
            gn = torch.tensor(im0.shape)[[1, 0, 1, 0]]  # normalization gain whwh
            if det is not None and len(det):
                # Rescale boxes from img_size to im0 size
                det[:, :4] = scale_coords(img.shape[2:], det[:, :4], im0.shape).round()

                # Print results
                for c in det[:, 5].unique():
                    n = (det[:, 5] == c).sum()  # detections per class
                    s += '%g %ss, ' % (n, names[int(c)])  # add to string

                # Write results
                for de, lic_plat in zip(det, plat_num):
                    # xyxy,conf,cls,lic_plat=de[:4],de[4],de[5],de[6:]
                    *xyxy, conf, cls=de

                    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') as f:
                            f.write(('%g ' * 5 + '\n') % (cls, xywh))  # label format

                    if save_img or view_img:  # Add bbox to image
                        # label = '%s %.2f' % (names[int(cls)], conf)
                        lb = ""
                        for a,i in enumerate(lic_plat):
                            # if a ==0:
                            #     continue
                            lb += CHARS[int(i)]
                        label = '%s %.2f' % (lb, conf)
                        im0=plot_one_box(xyxy, im0, label=label, color=colors[int(cls)], line_thickness=3)

            # Print time (demo + NMS)
            print('%sDone. (%.3fs)' % (s, t2 - t1))

            # Stream results
            if view_img:
                cv2.imshow(p, im0)
                if cv2.waitKey(1) == ord('q'):  # q to quit
                    raise StopIteration

            # Save results (image with detections)
            if save_img:
                if dataset.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

                        fourcc = 'mp4v'  # rec_result video codec
                        fps = vid_cap.get(cv2.CAP_PROP_FPS)
                        w = int(vid_cap.get(cv2.CAP_PROP_FRAME_WIDTH))
                        h = int(vid_cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
                        vid_writer = cv2.VideoWriter(save_path, cv2.VideoWriter_fourcc(*fourcc), fps, (w, h))
                    vid_writer.write(im0)

    if save_txt or save_img:
        print('Results saved to %s' % os.getcwd() + os.sep + out)
        if platform == 'darwin':  # MacOS
            os.system('open ' + save_path)

    print('Done. (%.3fs)' % (time.time() - t0))


if __name__ == '__main__':
    parser = argparse.ArgumentParser()
    parser.add_argument('--classify', nargs='+', type=str, default=True, help='True rec')
    parser.add_argument('--det-weights', nargs='+', type=str, default='./weights/yolov5_best.pt', help='model.pt path(s)')
    parser.add_argument('--rec-weights', nargs='+', type=str, default='./weights/lprnet_best.pth', help='model.pt path(s)')
    parser.add_argument('--source', type=str, default='./demo/images/', help='source')  # file/folder, 0 for webcam
    parser.add_argument('--rec_result', type=str, default='demo/rec_result', help='rec_result folder')  # rec_result folder
    parser.add_argument('--img-size', type=int, default=640, help='demo size (pixels)')
    parser.add_argument('--conf-thres', type=float, default=0.4, help='object confidence threshold')
    parser.add_argument('--iou-thres', type=float, default=0.5, help='IOU threshold for NMS')
    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
    parser.add_argument('--view-img', action='store_true', help='display results')
    parser.add_argument('--save-txt', action='store_true', help='save results to *.txt')
    parser.add_argument('--classes', nargs='+', type=int, help='filter by class')
    parser.add_argument('--agnostic-nms', action='store_true', help='class-agnostic NMS')
    parser.add_argument('--augment', action='store_true', help='augmented demo')
    parser.add_argument('--update', action='store_true', help='update all models')
    opt = parser.parse_args()
    print(opt)

    with torch.no_grad():
        if opt.update:  # update all models (to fix SourceChangeWarning)
            for opt.weights in ['yolov5s.pt', 'yolov5m.pt', 'yolov5l.pt', 'yolov5x.pt', 'yolov3-spp.pt']:
                detect()
                create_pretrained(opt.weights, opt.weights)
        else:
            detect()

Reference

CSDN: linux-mobaxterm-yolov5训练数据集ccpd–无数踩雷后

Github: https://github.com/ultralytics/yolov5

Github: https://github.com/sirius-ai/LPRNet_Pytorch

Gitee: https://gitee.com/reason1251326862/plate_classification
Github:https://github.com/kiloGrand/License-Plate-Recognition