一、源码下载

可以通过下方链接下载Efficientdet源码

GitHub - zylo117/Yet-Another-EfficientDet-Pytorch: The pytorch re-implement of the official efficientdet with SOTA performance in real time and pretrained weights.The pytorch re-implement of the official efficientdet with SOTA performance in real time and pretrained weights. - zylo117/Yet-Another-EfficientDet-Pytorchhttps://github.com/zylo117/Yet-Another-EfficientDet-Pytorch/tree/master

二、环境配置

1、使用anaconda创建环境

conda create -n efficient python==3.8 -y

2、进入环境

conda activate efficient

3、安装依赖包

pip install pycocotools-windows numpy opencv-python tqdm tensorboard tensorboardX pyyaml webcolors

4、安装torch和torchvision

pip install torch==1.9.1+cu111 torchvision==0.10.1+cu111 -f https://download.pytorch.org/whl/torch_stable.html

5、测试配置

5.1预训练权重下载

在源码下载的地方将预训练权重下载下来，这里选择的是Efficientdet-D1版本的预训练权重

在根目录下创建weights文件夹，将预训练权重放在文件夹内

5.2修改参数

在efficientdet_test.py中修改测试图片路径和efficientdet的版本，这里我下的版本是D1，所以就要将compound_coef改为1，如果你下的是其他版本，就将compound_coef改成对应的版本数字。

测试图片官方已经放在test\img.png，修改完之后运行efficientdet_test.py，检测结果保存在test文件夹内。

至此，efficientdet的所有环境配置结束。

三、自己的数据集制作

由于efficientdet使用的是coco格式的数据集，需要将标签转为coco格式，如果使用labelme进行标注，可以直接在labelme中切换成json格式的标签，最后用脚本进行合并，本文介绍的是使用VOC格式（xml）如何转为coco格式数据集。

1、划分验证集和训练集

使用下面的脚本将训练集进行划分：

# 将标签格式为xml的数据集按照8：2的比例划分为训练集和验证集

import os
import shutil
import random
from tqdm import tqdm


def split_img(img_path, label_path, split_list):
    try:  # 创建数据集文件夹
        Data = 'yourdatasetsname'
        os.mkdir(Data)

        train_img_dir = Data + '/images/train'
        val_img_dir = Data + '/images/val'
        # test_img_dir = Data + '/images/test'

        train_label_dir = Data + '/labels/train'
        val_label_dir = Data + '/labels/val'
        # test_label_dir = Data + '/labels/test'

        # 创建文件夹
        os.makedirs(train_img_dir)
        os.makedirs(train_label_dir)
        os.makedirs(val_img_dir)
        os.makedirs(val_label_dir)
        # os.makedirs(test_img_dir)
        # os.makedirs(test_label_dir)

    except:
        print('文件目录已存在')

    train, val = split_list
    all_img = os.listdir(img_path)
    all_img_path = [os.path.join(img_path, img) for img in all_img]
    # all_label = os.listdir(label_path)
    # all_label_path = [os.path.join(label_path, label) for label in all_label]
    train_img = random.sample(all_img_path, int(train * len(all_img_path)))
    train_img_copy = [os.path.join(train_img_dir, img.split('\\')[-1]) for img in train_img]
    train_label = [toLabelPath(img, label_path) for img in train_img]
    train_label_copy = [os.path.join(train_label_dir, label.split('\\')[-1]) for label in train_label]
    for i in tqdm(range(len(train_img)), desc='train2007 ', ncols=80, unit='img'):
        _copy(train_img[i], train_img_dir)
        _copy(train_label[i], train_label_dir)
        all_img_path.remove(train_img[i])
    val_img = all_img_path
    val_label = [toLabelPath(img, label_path) for img in val_img]
    for i in tqdm(range(len(val_img)), desc='test2007 ', ncols=80, unit='img'):
        _copy(val_img[i], val_img_dir)
        _copy(val_label[i], val_label_dir)


def _copy(from_path, to_path):
    shutil.copy(from_path, to_path)


def toLabelPath(img_path, label_path):
    img = img_path.split('\\')[-1]
    label = img.split('.jpg')[0] + '.xml'
    return os.path.join(label_path, label)


def main():
    img_path = '/path/to/your/image/folder'
    label_path = '/path/to/your/xml/folder'
    split_list = [0.8, 0.2]  # 数据集划分比例[train2007:test2007]
    split_img(img_path, label_path, split_list)


if __name__ == '__main__':
    main()

2、移动图片

在数据集根目录下创建datasets文件夹，并创建coco文件夹，在coco文件夹内创建annotations、train2017和val2017文件夹。将上一步划分好的训练集图片放入trian2017文件夹内，验证集图片放入val2017文件夹内。

3、生成json文件

使用下方的脚本分别对验证集和训练集的所有xml文件转为json文件，结果会生成两个json文件。

import xml.etree.ElementTree as ET
import os
import json

coco = dict()
coco['images'] = []
coco['type'] = 'instances'
coco['annotations'] = []
coco['categories'] = []

category_set = dict()
image_set = set()

# category_item_id = -1
# VOC数据集的类别id与coco数据集一样都是从1开始，如果初始设为-1，那么转出来的coco的json文件中category_id和类别id会从0开始，不符合coco标准，在调coco.py的时候会报错list越界
category_item_id = 0
image_id = 20180000000
annotation_id = 0

def addCatItem(name):
    global category_item_id
    category_item = dict()
    category_item['supercategory'] = 'none'
    category_item_id += 1
    category_item['id'] = category_item_id
    category_item['name'] = name
    coco['categories'].append(category_item)
    category_set[name] = category_item_id
    return category_item_id

def addImgItem(file_name, size):
    global image_id
    if file_name is None:
        raise Exception('Could not find filename tag in xml file.')
    if size['width'] is None:
        raise Exception('Could not find width tag in xml file.')
    if size['height'] is None:
        raise Exception('Could not find height tag in xml file.')
    image_id += 1
    image_item = dict()
    image_item['id'] = image_id
    image_item['file_name'] = file_name
    image_item['width'] = size['width']
    image_item['height'] = size['height']
    coco['images'].append(image_item)
    image_set.add(file_name)
    return image_id

def addAnnoItem(object_name, image_id, category_id, bbox):
    global annotation_id
    annotation_item = dict()
    annotation_item['segmentation'] = []
    seg = []
    # bbox[] is x,y,w,h
    # left_top
    seg.append(bbox[0])
    seg.append(bbox[1])
    # left_bottom
    seg.append(bbox[0])
    seg.append(bbox[1] + bbox[3])
    # right_bottom
    seg.append(bbox[0] + bbox[2])
    seg.append(bbox[1] + bbox[3])
    # right_top
    seg.append(bbox[0] + bbox[2])
    seg.append(bbox[1])

    annotation_item['segmentation'].append(seg)

    annotation_item['area'] = bbox[2] * bbox[3]
    annotation_item['iscrowd'] = 0
    annotation_item['ignore'] = 0
    annotation_item['image_id'] = image_id
    annotation_item['bbox'] = bbox
    annotation_item['category_id'] = category_id
    annotation_id += 1
    annotation_item['id'] = annotation_id
    coco['annotations'].append(annotation_item)

def _read_image_ids(image_sets_file):
    ids = []
    with open(image_sets_file) as f:
        for line in f:
            ids.append(line.rstrip())
    return ids

"""通过txt文件生成"""
#split ='train' 'va' 'trainval' 'test'
def parseXmlFiles_by_txt(data_dir,json_save_path,split='train'):
    print("hello")
    labelfile=split+".txt"
    image_sets_file = data_dir + "/ImageSets/Main/"+labelfile
    ids=_read_image_ids(image_sets_file)

    for _id in ids:
        xml_file=data_dir + f"/Annotations/{_id}.xml"

        bndbox = dict()
        size = dict()
        current_image_id = None
        current_category_id = None
        file_name = None
        size['width'] = None
        size['height'] = None
        size['depth'] = None

        tree = ET.parse(xml_file)
        root = tree.getroot()
        if root.tag != 'annotation':
            raise Exception('pascal voc xml root element should be annotation, rather than {}'.format(root.tag))

        # elem is <folder>, <filename>, <size>, <object>
        for elem in root:
            current_parent = elem.tag
            current_sub = None
            object_name = None

            if elem.tag == 'folder':
                continue

            if elem.tag == 'filename':
                file_name = elem.text
                if file_name in category_set:
                    raise Exception('file_name duplicated')

            # add img item only after parse <size> tag
            elif current_image_id is None and file_name is not None and size['width'] is not None:
                if file_name not in image_set:
                    current_image_id = addImgItem(file_name, size)
                    print('add image with {} and {}'.format(file_name, size))
                else:
                    raise Exception('duplicated image: {}'.format(file_name))
                    # subelem is <width>, <height>, <depth>, <name>, <bndbox>
            for subelem in elem:
                bndbox['xmin'] = None
                bndbox['xmax'] = None
                bndbox['ymin'] = None
                bndbox['ymax'] = None

                current_sub = subelem.tag
                if current_parent == 'object' and subelem.tag == 'name':
                    object_name = subelem.text
                    if object_name not in category_set:
                        current_category_id = addCatItem(object_name)
                    else:
                        current_category_id = category_set[object_name]

                elif current_parent == 'size':
                    if size[subelem.tag] is not None:
                        raise Exception('xml structure broken at size tag.')
                    size[subelem.tag] = int(subelem.text)

                # option is <xmin>, <ymin>, <xmax>, <ymax>, when subelem is <bndbox>
                for option in subelem:
                    if current_sub == 'bndbox':
                        if bndbox[option.tag] is not None:
                            raise Exception('xml structure corrupted at bndbox tag.')
                        bndbox[option.tag] = int(option.text)

                # only after parse the <object> tag
                if bndbox['xmin'] is not None:
                    if object_name is None:
                        raise Exception('xml structure broken at bndbox tag')
                    if current_image_id is None:
                        raise Exception('xml structure broken at bndbox tag')
                    if current_category_id is None:
                        raise Exception('xml structure broken at bndbox tag')
                    bbox = []
                    # x
                    bbox.append(bndbox['xmin'])
                    # y
                    bbox.append(bndbox['ymin'])
                    # w
                    bbox.append(bndbox['xmax'] - bndbox['xmin'])
                    # h
                    bbox.append(bndbox['ymax'] - bndbox['ymin'])
                    print('add annotation with {},{},{},{}'.format(object_name, current_image_id, current_category_id,
                                                                   bbox))
                    addAnnoItem(object_name, current_image_id, current_category_id, bbox)
    json.dump(coco, open(json_save_path, 'w'))

"""直接从xml文件夹中生成"""
def parseXmlFiles(xml_path,json_save_path):
    for f in os.listdir(xml_path):
        if not f.endswith('.xml'):
            continue

        # print(path)

        bndbox = dict()
        size = dict()
        current_image_id = None
        current_category_id = None
        file_name = None
        size['width'] = None
        size['height'] = None
        size['depth'] = None

        xml_file = os.path.join(xml_path, f)
        print(xml_file)

        tree = ET.parse(xml_file)
        root = tree.getroot()
        if root.tag != 'annotation':
            raise Exception('pascal voc xml root element should be annotation, rather than {}'.format(root.tag))

        # elem is <folder>, <filename>, <size>, <object>
        for elem in root:
            current_parent = elem.tag
            current_sub = None
            object_name = None

            if elem.tag == 'folder':
                continue

            if elem.tag == 'filename':
                # 改成与xml文件同名的图片文件
                file_name = f.split(".")[0] + ".jpg"
                # file_name = elem.text
                # if file_name in category_set:
                #     raise Exception('file_name duplicated')

            # add img item only after parse <size> tag
            elif current_image_id is None and file_name is not None and size['width'] is not None:
                if file_name not in image_set:
                    current_image_id = addImgItem(file_name, size)
                    print('add image with {} and {}'.format(file_name, size))
                else:
                    raise Exception('duplicated image: {}'.format(file_name))
                    # subelem is <width>, <height>, <depth>, <name>, <bndbox>
            for subelem in elem:
                bndbox['xmin'] = None
                bndbox['xmax'] = None
                bndbox['ymin'] = None
                bndbox['ymax'] = None

                current_sub = subelem.tag
                if current_parent == 'object' and subelem.tag == 'name':
                    object_name = subelem.text
                    if object_name not in category_set:
                        current_category_id = addCatItem(object_name)
                    else:
                        current_category_id = category_set[object_name]

                elif current_parent == 'size':
                    if size[subelem.tag] is not None:
                        raise Exception('xml structure broken at size tag.')
                    size[subelem.tag] = int(subelem.text)

                # option is <xmin>, <ymin>, <xmax>, <ymax>, when subelem is <bndbox>
                for option in subelem:
                    if current_sub == 'bndbox':
                        if bndbox[option.tag] is not None:
                            raise Exception('xml structure corrupted at bndbox tag.')
                        bndbox[option.tag] = int(option.text)

                # only after parse the <object> tag
                if bndbox['xmin'] is not None:
                    if object_name is None:
                        raise Exception('xml structure broken at bndbox tag')
                    if current_image_id is None:
                        raise Exception('xml structure broken at bndbox tag')
                    if current_category_id is None:
                        raise Exception('xml structure broken at bndbox tag')
                    bbox = []
                    # x
                    bbox.append(bndbox['xmin'])
                    # y
                    bbox.append(bndbox['ymin'])
                    # w
                    bbox.append(bndbox['xmax'] - bndbox['xmin'])
                    # h
                    bbox.append(bndbox['ymax'] - bndbox['ymin'])
                    print('add annotation with {},{},{},{}'.format(object_name, current_image_id, current_category_id,
                                                                   bbox))
                    addAnnoItem(object_name, current_image_id, current_category_id, bbox)
    json.dump(coco, open(json_save_path, 'w'))

if __name__ == '__main__':
    #通过txt文件生成
    # voc_data_dir="E:/VOCdevkit/VOC2007"
    # json_save_path="E:/VOCdevkit/voc2007trainval.json"
    # parseXmlFiles_by_txt(voc_data_dir,json_save_path,"trainval")

    #通过文件夹生成
    ann_path='/path/to/your/xml/folder'
    json_save_path="instances_val.json"
    parseXmlFiles(ann_path,json_save_path)

四、训练

1、创建配置文件

类似于YOLO，efficientdet也需要创建配置文件，直接复制projects文件夹的coco.yml并改名，并根据自己的数据集进行修改。

2、修改训练参数

3、训练

使用下方命令进行训练

python train.py