Python深度學習7：車牌檢測

► 前言

現在許多停車場使用車牌識別系統，停車時需要先使用攝像頭拍攝影像，透過AI檢測車牌再進行識別，本篇博文介紹如何訓練車牌檢測模型。

► 程式碼講解

在Kaggle網站內搜尋Car License Plate Detection並下載，解壓縮資料夾，更改名稱為“CarLicensePlateDetection”，將資料夾上傳至Google雲端硬碟，開啟Colab並設定為GPU執行，連結Google雲段硬碟。

先建立訓練資料夾，程式碼如下：

import numpy as np
import pandas as pd
import os
import numpy

os.mkdir("/content/plate_train_data")
os.mkdir("/content/plate_train_data/images")
os.mkdir("/content/plate_train_data/labels")
os.mkdir("/content/plate_train_data/images/train")
os.mkdir("/content/plate_train_data/images/val")
os.mkdir("/content/plate_train_data/labels/train")
os.mkdir("/content/plate_train_data/labels/val")

將 xml 文件轉換為 txt 文件，因為 Yolo需要使用歸一化邊界框的 txt 文件，格式為「class id, x, y, width, height」，資料呈現如下：

0 0.39375 0.34615384615384615 0.0975 0.05263157894736842

0 0.64375 0.35020242914979755 0.0625 0.06072874493927125

轉換程式碼如下：

import xml.etree.ElementTree as ET
from xml.dom.minidom import parse

path = "/content/drive/MyDrive/AI/dataset/CarLicensePlateDetection/annotations"
classes = {"licence":0}

for annotations in os.listdir(path):
    dom = parse(os.path.join(path,annotations))
    root = dom.documentElement
    filename = ".txt".join(root.getElementsByTagName("filename")[0].childNodes[0].data.split(".png"))
    image_width = root.getElementsByTagName("width")[0].childNodes[0].data
    image_height = root.getElementsByTagName("height")[0].childNodes[0].data
    with open("/content/plate_train_data/labels/train/"+filename,"w") as r:
        for items in root.getElementsByTagName("object") :
            name = items.getElementsByTagName("name")[0].childNodes[0].data
            xmin = items.getElementsByTagName("xmin")[0].childNodes[0].data
            ymin = items.getElementsByTagName("ymin")[0].childNodes[0].data
            xmax = items.getElementsByTagName("xmax")[0].childNodes[0].data
            ymax = items.getElementsByTagName("ymax")[0].childNodes[0].data
            x_center_norm = ((int(xmin)+int(xmax)) / 2 ) / int(image_width)
            y_center_norm = ((int(ymin)+int(ymax))/2) / int(image_height)
            width_norm = ((int(xmax)-int(xmin))/int(image_width))
            height_norm = ((int(ymax)-int(ymin))/int(image_height))
            r.write(str(classes[name])+" ")
            r.write(str(x_center_norm)+" ")
            r.write(str(y_center_norm)+" ")
            r.write(str(width_norm)+" ")
            r.write(str(height_norm)+"\n")

複製CarLicensePlateDetection內的image至plate_train_data/images/train

import shutil
path = "/content/drive/MyDrive/AI/dataset/CarLicensePlateDetection/images"

將訓練資料集亂數分割120至驗證資料集內

from random import shuffle
from glob import glob
import random
import shutil

random.seed(42)
files = glob("/content/plate_train_data/images/train/*.png")
shuffle(files)
txt_train_path = "/content/plate_train_data/labels/train/"
img_train_path = "/content/plate_train_data/images/train/"
txt_val_path = "/content/plate_train_data/labels/val/"
img_val_path = "/content/plate_train_data/images/val/"
for i in files[:120]:
    filename = i.split("/")[-1]
    shutil.move(txt_train_path+filename.replace(".png",".txt"),\
                txt_val_path+filename.replace(".png",".txt"))
    shutil.move(img_train_path+filename,img_val_path+filename)

建立yaml檔案，train後面為train資料集位置

%cd /content
!echo "train: /content/plate_train_data/images/train" > licence.yaml
!echo "val:   /content/plate_train_data/images/val"  >> licence.yaml
!echo "nc : 1" >> licence.yaml
!echo "names: ['licence']" >> licence.yaml

下載Yolov5，並安裝相關套件

%cd /content/
%rm -rf yolov5
!git clone https://github.com/ultralytics/yolov5
%cd yolov5
!pip install -r requirements.txt

開始訓練

訓練指令如下：

!python train.py --weights yolov5s.pt --cfg models/yolov5s.yaml --img 416 --batch 48 --epochs 40 --data /content/licence.yaml  --cache

訓練完成後顯示推論結果

from IPython.display import Image
Image('/content/yolov5/runs/train/exp/val_batch0_labels.jpg')

►小結

以上流程是使用Yolov5訓練車牌檢測的流程，當檢測出車牌就可以裁切進行文字OCR，讓客戶繳費時輸入自己的車牌，點選自己的車子照片進行繳費，之後有空再來整理發文。

► 參考資料

Car License Plate Detection