you only look once 就能懂
1. 转换标注文件：（xml----->txt） in_file = open(‘VOCdevkit/VOC%s/Annotations/%s.xml’%(year, image_id)) 打开xml文件 tree=ET.parse(in_file) 解将xml文件析成ElementTree类的对象 root = tree.getroot() 获取xml文件的根节点
接下来是一个循环： for obj in root.iter(‘object’): 对于根节点中的条目 object的循环： 由于我这里只有一个目标，所以应该只是循环一次
difficult = obj.find(‘difficult’).text 应该指图像中是否有检测目标
cls = obj.find(‘name’).text 这里是people
cls_id = classes.index(cls) 之前定义过 classes = [“people”] 这里获取索引（第几类检测对象）
xmlbox = obj.find(‘bndbox’) 找到boundingbox的边界的父条目
b = (int(xmlbox.find(‘xmin’).text), int(xmlbox.find(‘ymin’).text), int(xmlbox.find(‘xmax’).text), int(xmlbox.find(‘ymax’).text)) 将父条目下的边界数值保存为元组b
list_file.write(" " + “,”.join([str(a) for a in b]) + ‘,’ + str(cls_id)) 说明了下面的五个数字都是什么 分别是四个边界值 + 检测对象类别
def convert_annotation(year, image_id, list_file): in_file = open('VOCdevkit/VOC%s/Annotations/%s.xml'%(year, image_id)) tree=ET.parse(in_file) root = tree.getroot() for obj in root.iter('object'): difficult = obj.find('difficult').text cls = obj.find('name').text if cls not in classes or int(difficult)==1: continue cls_id = classes.index(cls) xmlbox = obj.find('bndbox') b = (int(xmlbox.find('xmin').text), int(xmlbox.find('ymin').text), int(xmlbox.find('xmax').text), int(xmlbox.find('ymax').text)) list_file.write(" " + ",".join([str(a) for a in b]) + ',' + str(cls_id))2. wd = getcwd() 获得此py文件所获得的目录
3. 下面进行一个循环
对于年份、什么集合（训练、验证、测试，因为~VOCdevkit\VOC2012\ImageSets\Main下分为训练、验证、测试集几个txt，txt里面写着哪些图片用来训练，那些用来验证，那些用来测试）进行循环：
.strip().split() .strip() 删除字符串头尾的空白符（包括’\n’, ‘\r’, ‘\t’, ’ ') .split() 以空字符进行字符串分割
list_file = open(’%s_%s.txt’%(year, image_set), ‘w’) 创建好txt文件准备写入需要的信息
convert_annotation(year, image_id, list_file) 将xml的信息提取出来写入项目文件夹下的txt文件，如2012_test.txt
for year, image_set in sets: image_ids = open('VOCdevkit/VOC%s/ImageSets/Main/%s.txt'%(year, image_set)).read().strip().split() list_file = open('%s_%s.txt'%(year, image_set), 'w') for image_id in image_ids: list_file.write('%s/VOCdevkit/VOC%s/JPEGImages/%s.jpg'%(wd, year, image_id)) convert_annotation(year, image_id, list_file) list_file.write('\n') list_file.close()下面是训练代码
导入必要的包
import numpy as np import keras.backend as K from keras.layers import Input, Lambda from keras.models import Model from keras.callbacks import TensorBoard, ModelCheckpoint, EarlyStopping from yolo3.model import preprocess_true_boxes, yolo_body, tiny_yolo_body, yolo_loss from yolo3.utils import get_random_data import tensorflow as tf import keras.backend.tensorflow_backend as KTF定义主函数
annotation_path为由voc_annotation.py生成的txt文件（再说一遍，因为xml并不是yolo所识别的数据格式，因此需要用这个文件将xml文件转换为txt信息。） 打开’2012_train.txt’，我们可以看到： ~/train1.jpg 189,171,233,299,0 后面有五个数字：189,171,233,299,0 分别表示 什么呢？ 之前讲过，前四个数字约束了boundingbox的大小，0代表了第0类别
log_dir = ‘logs/000/’ 模型保存地址 classes_path = ‘model_data/voc_classes.txt’ 内容为people（这里我仅用了一类）
anchors_path = ‘model_data/yolo_anchors.txt’ yolo中的anchor到底是什么呢？ 直接上吴恩达的教程 https://mooc.study.163.com/learn/2001281004?tid=2001392030#/learn/content?type=detail&id=2001729339 简而言之，anchor box是为了处理一个格子中有多个检测对象的情况，一般情况下anchorbox可以设置5-10个，涵盖想要检测的各种对象的形状。打开yolo_anchors文件就可以看到内容为： 10,13, 16,30, 33,23, 30,61, 62,45, 59,119, 116,90, 156,198, 373,326 九个anchorbox
input_shape = (416, 416) 必须与输入一致（yolo算法不能改变输入图像的尺寸）
def _main(): annotation_path = '2012_train.txt' log_dir = 'logs/000/' classes_path = 'model_data/voc_classes.txt' anchors_path = 'model_data/yolo_anchors.txt' class_names = get_classes(classes_path) anchors = get_anchors(anchors_path) input_shape = (416, 416) # multiple of 32, hw model = create_model(input_shape, anchors, len(class_names) ) train(model, annotation_path, input_shape, anchors, len(class_names), log_dir=log_dir)下面分别看主函数中嵌套的几个函数：
获得类别函数get_class:f.readlines()读取整个文件，然后把每一行放到一个列表里。所以你要是向voc_classes文件中添加类别时候一定要另起一行添加，如： 返回一个由各个类别名称组成的列表
def get_classes(classes_path): with open(classes_path) as f: class_names = f.readlines() class_names = [c.strip() for c in class_names] return class_names 获得anchorbox函数get_anchorsf.readline()读取一行,是在上一次读取的基础上读取。在这里只是读取一行用逗号分隔 所以在添加a时需要全部用逗号隔开。
最后需要reshape成为一个（9， 2 ）的形式（9表示9个box，2表示一个box的长宽）
def get_anchors(anchors_path): with open(anchors_path) as f: anchors = f.readline() anchors = [float(x) for x in anchors.split(',')] return np.array(anchors).reshape(-1, 2) create_model 建立模型函数image_input = Input(shape=(None, None, 3))
y_true = [Input(shape=(h//{0:32, 1:16, 2:8}[l], w//{0:32, 1:16, 2:8}[l], num_anchors//3,num_classes+5)) for l in range(3)] h = w = 416 num_anchors为anchorbox的个数 num_classes为类别个数 为什么y_true是这样呢？存疑
from yolo3.model import preprocess_true_boxes, yolo_body, tiny_yolo_body, yolo_loss导入了yolobody，因此这里直接获取yolo结构
model_body = yolo_body(image_input, num_anchors//3, num_classes) 至于为什么要地板除3还是不明白
这个model_body到底是什么样子呢？ 一共252层，最后7层可训练
model = Model([model_body.input, *y_true], model_loss) 指定输入是什么输出是什么，在跳远连接的代码示例中有用到
def create_model(input_shape, anchors, num_classes, load_pretrained=False, freeze_body=False, weights_path='model_data/yolo_weights.h5'): K.clear_session() # get a new session image_input = Input(shape=(None, None, 3)) h, w = input_shape num_anchors = len(anchors) y_true = [Input(shape=(h//{0:32, 1:16, 2:8}[l], w//{0:32, 1:16, 2:8}[l], \ num_anchors//3, num_classes+5)) for l in range(3)] model_body = yolo_body(image_input, num_anchors//3, num_classes) print('Create YOLOv3 model with {} anchors and {} classes.'.format(num_anchors, num_classes)) if load_pretrained: model_body.load_weights(weights_path, by_name=True, skip_mismatch=True) print('Load weights {}.'.format(weights_path)) if freeze_body: # Do not freeze 3 output layers. num = len(model_body.layers)-7 for i in range(num): model_body.layers[i].trainable = False print('Freeze the first {} layers of total {} layers.'.format(num, len(model_body.layers))) model_loss = Lambda(yolo_loss, output_shape=(1,), name='yolo_loss', arguments={'anchors': anchors, 'num_classes': num_classes, 'ignore_thresh': 0.5})( [*model_body.output, *y_true]) model = Model([model_body.input, *y_true], model_loss) return model 训练函数 train(model, annotation_path, input_shape, anchors, len(class_names), log_dir=log_dir) 嵌套了 data_generator_wrapdef data_generator(annotation_lines, batch_size, input_shape, anchors, num_classes): n = len(annotation_lines) np.random.shuffle(annotation_lines) i = 0 while True: image_data = [] box_data = [] for b in range(batch_size): i %= n image, box = get_random_data(annotation_lines[i], input_shape, random=True) image_data.append(image) box_data.append(box) i += 1 image_data = np.array(image_data) box_data = np.array(box_data) y_true = preprocess_true_boxes(box_data, input_shape, anchors, num_classes) yield [image_data, *y_true], np.zeros(batch_size)
def train(model, annotation_path, input_shape, anchors, num_classes, log_dir='logs/'): model.compile(optimizer='adam', loss={ 'yolo_loss': lambda y_true, y_pred: y_pred}) logging = TensorBoard(log_dir=log_dir) checkpoint = ModelCheckpoint(log_dir + "ep{epoch:03d}-loss{loss:.3f}-val_loss{val_loss:.3f}.h5", monitor='val_loss', save_weights_only=True, save_best_only=True, period=1) batch_size = 3 val_split = 0.1 with open(annotation_path) as f: lines = f.readlines() np.random.shuffle(lines) num_val = int(len(lines)*val_split) num_train = len(lines) - num_val print('Train on {} samples, val on {} samples, with batch size {}.'.format(num_train, num_val, batch_size)) model.fit_generator(data_generator_wrap(lines[:num_train], batch_size, input_shape, anchors, num_classes), steps_per_epoch=max(1, num_train//batch_size), validation_data=data_generator_wrap(lines[num_train:], batch_size, input_shape, anchors, num_classes), validation_steps=max(1, num_val//batch_size), epochs=100, # 500 有点多了 initial_epoch=0) model.save_weights(log_dir + 'trained_weights.h5')剩下的可以看这一篇博文
