1. Anylabeling自动标记

使用以下代码

import numpy as np
import torch
import matplotlib.pyplot as plt
import cv2
import json
import sys
import os
from segment_anything import sam_model_registry, SamAutomaticMaskGenerator

def segment(imgPath):
    # 通过opencv图取图像
    print("开始读图像")
    image = cv2.imread(imgPath)
    image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
    # 添加当前系统路径，添加模型文件路径
    sys.path.append("..")
    sam_checkpoint = "sam_vit_l_0b3195.pth"
    model_type = "vit_l"
    #设置运行推理的设备
    device = "cuda"
    # 创建sam模型推理对象
    sam = sam_model_registry[model_type](checkpoint=sam_checkpoint)
    sam.to(device=device)
    mask_generator = SamAutomaticMaskGenerator(sam)
    masks = mask_generator.generate(image)
    # 给分割出来的物体上色，显示分割效果
    show_anns(masks, imgPath)
    torch.cuda.empty_cache()
    
def show_anns(anns, imgPath):
    if len(anns) == 0:
        return
    # 对检测结果的字典对象进行排序
    sorted_anns = sorted(anns, key=(lambda x: x['area']), reverse=True)
    ax = plt.gca()
    ax.set_autoscale_on(False)
    img = np.ones((sorted_anns[0]['segmentation'].shape[0], sorted_anns[0]['segmentation'].shape[1], 4))
    img[:,:,3] = 0
    shapes = []
    for ann in sorted_anns:
        # 过滤面积比较小的物体
        if ann['area'] >=10:
            # 创建labelme格式
            tempData = {"label": "air",  ###### 需要修改标签名 ###########
                        "points": [],
                        "group_id": None,
                        "shape_type": "polygon",
                        "flags": {}
                        }
            # 获取分割物体掩膜
            m = ann['segmentation']

            # 找出物体轮廓
            objImg = np.zeros((m.shape[0], m.shape[1], 1), np.uint8)
            objImg[m] = 255
            contours, hierarchy = cv2.findContours(objImg, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

            # 找出轮廓最大的
            max_area = 0
            maxIndex = 0
            for i in range(0, len(contours)):
                area = cv2.contourArea(contours[i])
                if area >= max_area:
                    max_area = area
                    maxIndex = i

            # 将每个物体轮廓点数限制在一定范围内
            if len(contours[maxIndex]) >=30:
                contours = list(contours[maxIndex])
                contours = contours[::int(len(contours)/30)]
            else:
                contours = list(contours[maxIndex])
            # 向labelme数据格式中添加轮廓点
            for point in contours:
                tempData["points"].append([int(point[0][0]), int(point[0][1])])
            # 添加物体标注信息
            shapes.append(tempData)

            # 在彩色图像上标出物体
            color_mask = np.concatenate([np.random.random(3), [1]])
            img[m] = color_mask
    jsonPath = imgPath.replace(".bmp", ".json")  # 需要生成的文件路径
    print(jsonPath) #
    # 分离文件最后的名
    split_path = imgPath.split("//")
    title = split_path[-1]
    print("title为", title)
    # 创建json文件
    file_out = open(jsonPath, "w")
    # 载入json文件
    jsonData = {}
    # 写入，修改json文件
    jsonData["version"] = "5.2.1"
    jsonData["flags"] = {}
    jsonData["shapes"] = shapes
    jsonData["imagePath"] = title
    jsonData["imageData"] = None
    jsonData["imageHeight"] = sorted_anns[0]['segmentation'].shape[0]
    jsonData["imageWidth"] = sorted_anns[0]['segmentation'].shape[1]
    # 保存json文件
    file_out.write(json.dumps(jsonData, indent=4))  # 保存文件
    # 关闭json文件
    file_out.close()
    ax.imshow(img)

if __name__ == '__main__':
    # dir_path = r"20230922//aggregate//" #该py文件路径下
    dir_path = r"20230927//air//"            ###修改文件的路径####
   
    for image_name in os.listdir(dir_path):
        img_path = os.path.join(dir_path, image_name)
        print(img_path)
        # print(image_name)
        if img_path.endswith('.bmp'):
            segment(img_path)

2.使用Anylabeing或者labelme打开核对标记

1.用anylabeling打开

在Python的终端输入anylabeling

即可打开

模型加载不出来，可参考这篇博客。
打开文件夹之后如图所示。

但是发现删除图层的话，对于很小的图层会出现选不中的问题，可使用放大图片在选中的方式。但是创建图层很方便。

1.1 创建图层

2.用labelme打开

终端输入labelme,打开labelme工具，选择OpenDir,选择所对应的文件目录。



点击红色图层，按键盘上的delete可删除。会有部分图像没有图层，点击Create Polygons创建图层。
但是这种创建图层的方式又太麻烦，没有anylabeling创建图层的方式简单。

3.修改标签label名称(可选)

4. Jons文件格式转txt格式

标记结束后就会出现下述情况：图片和json文件在同一个文件夹。

将json文件转换为txt文件，使用以下代码

import os
import json
import numpy as np
# 类和索引
CLASSES=["brick","aggregate","air","other"]
def convert_2(size,box):
    w = (box[1] - box[0]) / size[0]
    h = (box[3] - box[2]) / size[1]
    x = (box[0]+box[1]) /2 /size[0]
    y = (box[2]+box[3]) /2 /size[1]
    return (x,y,w,h)

def json2txt(path_json,path_txt):
    with open(path_json,"r") as path_json:
        jsonx=json.load(path_json)
        width=int(jsonx["imageWidth"])      # 原图的宽
        height=int(jsonx["imageHeight"])    # 原图的高

        with open(path_txt,"w+") as ftxt:
            # 遍历每一个bbox对象
            for shape in jsonx["shapes"]:
                obj_cls=str(shape["label"])     # 获取类别
                cls_id=CLASSES.index(obj_cls)   # 获取类别索引

                points=np.array(shape["points"])    # 获取(x1,y1,x2,y2)
                #print(points)
                xmin = int(points[0][0])
                xmax = int(points[0][0])
                ymin = int(points[0][1])
                ymax = int(points[0][1])
                for point in shape["points"]:
                    if point[0] > xmax:
                        xmax = point[0]
                    if point[0] < xmin:
                        xmin = point[0]
                    if point[1] > ymax:
                        ymax = point[1]
                    if point[1] < ymin:
                        ymin = point[1]
                #print(xmin,xmax,ymin,ymax)
                # (左上角,右下角) -> (中心点,宽高) 归一化
                bb=convert_2((width,height),(xmin,xmax,ymin,ymax))

                ftxt.write(str(cls_id)+" "+" ".join([str(a) for a in bb])+"\n")#.join()
if __name__=="__main__":
    # json文件夹
    dir_json="E:\\dataset\\BrickConcreteDataSet\\20230927\\air\\"
    # txt文件夹
    dir_txt="E:\\dataset\BrickConcreteDataSet\\20230927\\lables_air\\"
    if not os.path.exists(dir_txt):
        os.makedirs(dir_txt)
    # 得到所有json文件
    list_json=os.listdir(dir_json)

    # 遍历每一个json文件,转成txt文件
    for cnt, json_name in enumerate(list_json):
        #print("cnt=%d,name=%s"%(cnt, json_name))
        file,fileext = os.path.splitext(json_name)
        if(fileext == ".json"):
            path_json = dir_json + json_name
            path_txt= dir_txt + json_name.replace(".json",".txt")
            # (x1,y1,x2,y2)->(x,y,w,h)
            json2txt(path_json,path_txt)
        else:
            pass

运行后可生成txt文件。

5. 按比例分数据集

将标记的数据集按比例分成训练集，验证集和测试集。
将图片和标签分别存放在不同的文件夹内。运行以下代码，可生成不同比例的数据集。

import os
import shutil
import random
random.seed(0)
def split_data(file_path,xml_path, new_file_path, train_rate, val_rate, test_rate):
    each_class_image = []
    each_class_label = []
    for image in os.listdir(file_path):
        each_class_image.append(image)
    for label in os.listdir(xml_path):
        each_class_label.append(label)
    data=list(zip(each_class_image,each_class_label))
    total = len(each_class_image)
    random.shuffle(data)
    each_class_image,each_class_label=zip(*data)
    train_images = each_class_image[0:int(train_rate * total)]
    val_images = each_class_image[int(train_rate * total):int((train_rate + val_rate) * total)]
    test_images = each_class_image[int((train_rate + val_rate) * total):]
    train_labels = each_class_label[0:int(train_rate * total)]
    val_labels = each_class_label[int(train_rate * total):int((train_rate + val_rate) * total)]
    test_labels = each_class_label[int((train_rate + val_rate) * total):]

    for image in train_images:
        print(image)
        old_path = file_path + '/' + image
        new_path1 = new_file_path + '/' + 'train' + '/' + 'images'
        if not os.path.exists(new_path1):
            os.makedirs(new_path1)
        new_path = new_path1 + '/' + image
        shutil.copy(old_path, new_path)

    for label in train_labels:
        print(label)
        old_path = xml_path + '/' + label
        new_path1 = new_file_path + '/' + 'train' + '/' + 'labels'
        if not os.path.exists(new_path1):
            os.makedirs(new_path1)
        new_path = new_path1 + '/' + label
        shutil.copy(old_path, new_path)

    for image in val_images:
        old_path = file_path + '/' + image
        new_path1 = new_file_path + '/' + 'val' + '/' + 'images'
        if not os.path.exists(new_path1):
            os.makedirs(new_path1)
        new_path = new_path1 + '/' + image
        shutil.copy(old_path, new_path)

    for label in val_labels:
        old_path = xml_path + '/' + label
        new_path1 = new_file_path + '/' + 'val' + '/' + 'labels'
        if not os.path.exists(new_path1):
            os.makedirs(new_path1)
        new_path = new_path1 + '/' + label
        shutil.copy(old_path, new_path)

    for image in test_images:
        old_path = file_path + '/' + image
        new_path1 = new_file_path + '/' + 'test' + '/' + 'images'
        if not os.path.exists(new_path1):
            os.makedirs(new_path1)
        new_path = new_path1 + '/' + image
        shutil.copy(old_path, new_path)

    for label in test_labels:
        old_path = xml_path + '/' + label
        new_path1 = new_file_path + '/' + 'test' + '/' + 'labels'
        if not os.path.exists(new_path1):
            os.makedirs(new_path1)
        new_path = new_path1 + '/' + label
        shutil.copy(old_path, new_path)


if __name__ == '__main__':
    file_path = "E:/VOCdata/image"//存放图片的文件路径
    xml_path = 'E:/VOCdata/labels'//存放标签的文件路径
    new_file_path = "E:/VOCdevkit"//生成数据集的路径
    split_data(file_path,xml_path, new_file_path, train_rate=0.7, val_rate=0.2, test_rate=0.1)