yolo v3 + tensorflow训练源码解析

1.训练源码地址

2.代码解析地址

"""
Retrain the YOLO model for your own dataset.
"""

import numpy as np
import keras.backend as K
from keras.layers import Input, Lambda
from keras.models import Model
from keras.optimizers import Adam
from keras.callbacks import TensorBoard, ModelCheckpoint, ReduceLROnPlateau, EarlyStopping

from yolo3.model import preprocess_true_boxes, yolo_body, tiny_yolo_body, yolo_loss
from yolo3.utils import get_random_data


def _main():
    annotation_path = '2028_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)
    num_classes = len(class_names)
    anchors = get_anchors(anchors_path)

    input_shape = (416,416) # multiple of 32, hw

    is_tiny_version = len(anchors)==6 # default setting
    if is_tiny_version:
        model = create_tiny_model(input_shape, anchors, num_classes,
            freeze_body=2, weights_path='model_data/tiny_yolo_weights.h5')
    else:
        model = create_model(input_shape, anchors, num_classes,
            freeze_body=2, weights_path='model_data/yolo_weights.h5') # make sure you know what you freeze
            # freeze_body=2, weights_path='model_data/yolo_weights.h5')
    logging = TensorBoard(log_dir=log_dir)

    # 1.
    # filename：字符串，保存模型的路径
    # 2.
    # monitor：需要监视的值，val_acc或这val_loss
    # 3.
    # verbose：信息展示模式，0
    # 为不打印输出信息，1
    # 打印
    # 4.
    # save_best_only：当设置为True时，将只保存在验证集上性能最好的模型
    # 5.
    # mode：‘auto’，‘min’，‘max’之一，在save_best_only = True时决定性能最佳模型的评判准则，例如，当监测值为val_acc时，模式应为max，当检测值为val_loss时，模式应为min。在auto模式下，评价准则由被监测值的名字自动推断。
    # 6.
    # save_weights_only：若设置为True，则只保存模型权重，否则将保存整个模型（包括模型结构，配置信息等）
    # 7.
    # period：CheckPoint之间的间隔的epoch数

    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=3)


    reduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.1, patience=3, verbose=1)
    early_stopping = EarlyStopping(monitor='val_loss', min_delta=0, patience=10, verbose=1)

    val_split = 0.1    #训练和验证的比例
    with open(annotation_path) as f:
        lines = f.readlines()
    np.random.seed(10101)
    np.random.shuffle(lines)
    np.random.seed(None)
    num_val = int(len(lines)*val_split)   #验证集数量
    num_train = len(lines) - num_val     #训练集数量

    # Train with frozen layers first, to get a stable loss.
    # Adjust num epochs to your dataset. This step is enough to obtain a not bad model.
    '''
    第1阶段，冻结部分网络，只训练底层权重。
    优化器使用最常见的Adam。
    损失函数，直接使用模型的输出y_pred，忽略真值y_true
    
    其中，关于损失函数yolo_loss，以及y_true和y_pred：

    把y_true当成输入，作为模型的多输入，把loss封装为层（即Lambda层），作为模型的输出；
    在模型中，最终输出的y_pred就是loss；
    在编译（compile）时，将loss设置为y_pred即可，无视y_true；
    在训练时，随意添加一个符合结构的y_true即可。
    '''

    if True:
        model.compile(optimizer=Adam(lr=1e-3), loss={
     
            # use custom yolo_loss Lambda layer.使用定制的 yolo_loss Lambda层
            'yolo_loss': lambda y_true, y_pred: y_pred})  #损失函数

        # 模型fit数据，使用数据生成包装器（data_generator_wrapper），按批次生成训练和验证数据。最终，模型model存储权重。实现如下：
        batch_size = 4
        print('Train on {} samples, val on {} samples, with batch size {}.'.format(num_train, num_val, batch_size))
        model.fit_generator(data_generator_wrapper(lines[:num_train], batch_size, input_shape, anchors, num_classes),
                steps_per_epoch=max(1, num_train//batch_size),
                validation_data=data_generator_wrapper(lines[num_train:], batch_size, input_shape, anchors, num_classes),
                validation_steps=max(1, num_val//batch_size),
                epochs=50,
                initial_epoch=0,
                callbacks=[logging, checkpoint])
        # 存储最终的去权重，再训练过程中，也通过回调存储
        model.save_weights(log_dir + 'trained_weights_stage_1.h5')

    # Unfreeze and continue training, to fine-tune.
    # Train longer if the result is not good.
    if True:
        for i in range(len(model.layers)):
            model.layers[i].trainable = True
        model.compile(optimizer=Adam(lr=1e-4), loss={
     'yolo_loss': lambda y_true, y_pred: y_pred}) # recompile to apply the change
        print('Unfreeze all of the layers.')


#需要调小点  32的batch size 1080ti 32G带不动
        '''
        第2阶段的模型fit数据，与第1阶段类似，从第50个epoch开始，一直训练到第100个epoch，当触发条件时，则提前终止。额外增加了两个回调reduce_lr和early_stopping，用于控制训练提取终止的时机：

reduce_lr：当评价指标不在提升时，减少学习率，每次减少10%（factor），当验证损失值（val_loss），持续3次未减少（patience）时，则终止训练。
early_stopping：当验证集损失值，连续增加小于0（min_delta）时，持续10个epoch（patience），则终止训练。
'''
        batch_size = 32 # note that more GPU memory is required after unfreezing the body
        print('Train on {} samples, val on {} samples, with batch size {}.'.format(num_train, num_val, batch_size))
        model.fit_generator(data_generator_wrapper(lines[:num_train], batch_size, input_shape, anchors, num_classes),
            steps_per_epoch=max(1, num_train//batch_size),
            validation_data=data_generator_wrapper(lines[num_train:], batch_size, input_shape, anchors, num_classes),
            validation_steps=max(1, num_val//batch_size),
            epochs=100,
            initial_epoch=50,
            callbacks=[logging, checkpoint, reduce_lr, early_stopping])
        model.save_weights(log_dir + 'trained_weights_final.h5')

    # Further training if needed.


def get_classes(classes_path):
    '''loads the classes'''
    with open(classes_path) as f:
        class_names = f.readlines()
    class_names = [c.strip() for c in class_names]
    return class_names

def get_anchors(anchors_path):
    '''loads the anchors from a file'''
    with open(anchors_path) as f:
        anchors = f.readline()
    anchors = [float(x) for x in anchors.split(',')]
    # print(np.array(anchors).reshape(-1, 2))
    '''
    [[ 10.  13.]
 [ 16.  30.]
 [ 33.  23.]
 [ 30.  61.]
 [ 62.  45.]
 [ 59. 119.]
 [116.  90.]
 [156. 198.]
 [373. 326.]]
    '''
    return np.array(anchors).reshape(-1, 2)



#freeze_body  1是冻结DarkNet53的层    2是冻结全部，只保留最后3层
#网络最后的3层：
#3个1x1的卷积层（代替全连接层），用于将3个尺度的特征图，转化为3个尺度的预测值
#weights_path  预训练权重文件
def create_model(input_shape, anchors, num_classes, load_pretrained=True, freeze_body=2,
            weights_path='model_data/yolo_weights.h5'):
    '''create the training model'''
    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 in [1, 2]:
            # Freeze darknet53 body or freeze all but 3 output layers.
            num = (185, len(model_body.layers)-3)[freeze_body-1]
            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

def create_tiny_model(input_shape, anchors, num_classes, load_pretrained=True, freeze_body=2,
            weights_path='model_data/tiny_yolo_weights.h5'):
    '''create the training model, for Tiny YOLOv3'''
    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}[l], w//{
     0:32, 1:16}[l], \
        num_anchors//2, num_classes+5)) for l in range(2)]

    model_body = tiny_yolo_body(image_input, num_anchors//2, num_classes)
    print('Create Tiny 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 in [1, 2]:
            # Freeze the darknet body or freeze all but 2 output layers.
            num = (20, len(model_body.layers)-2)[freeze_body-1]
            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': 0.7})(
        [*model_body.output, *y_true])
    model = Model([model_body.input, *y_true], model_loss)
    return model

def data_generator(annotation_lines, batch_size, input_shape, anchors, num_classes):
    '''data generator for fit_generator'''
    n = len(annotation_lines)
    i = 0
    while True:
        image_data = []
        box_data = []
        for b in range(batch_size):
            if i==0:
                np.random.shuffle(annotation_lines)   #随机抽取    shuffle-洗牌
            image, box = get_random_data(annotation_lines[i], input_shape, random=True)
            image_data.append(image)
            box_data.append(box)
            i = (i+1) % n
        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 data_generator_wrapper(annotation_lines, batch_size, input_shape, anchors, num_classes):
    n = len(annotation_lines)
    if n==0 or batch_size<=0: return None
    return data_generator(annotation_lines, batch_size, input_shape, anchors, num_classes)

if __name__ == '__main__':
    _main()