train.py

import numpy as np
import tensorflow as tf
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 nets.yolo4 import yolo_body
from nets.loss import yolo_loss
from keras.backend.tensorflow_backend import set_session
from utils.utils import get_random_data,get_random_data_with_Mosaic,rand,WarmUpCosineDecayScheduler


#---------------------------------------------------#
#   获得类和先验框
#---------------------------------------------------#
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(',')]
    return np.array(anchors).reshape(-1, 2)

#---------------------------------------------------#
#   训练数据生成器
#---------------------------------------------------#
def data_generator(annotation_lines, batch_size, input_shape, anchors, num_classes, mosaic=False):
    '''data generator for fit_generator'''
    n = len(annotation_lines)
    i = 0
    flag = True
    while True:
        image_data = []
        box_data = []
        for b in range(batch_size):
            if i==0:
                np.random.shuffle(annotation_lines)
            if mosaic:
                if flag and (i+4) < n:
                    image, box = get_random_data_with_Mosaic(annotation_lines[i:i+4], input_shape)
                    i = (i+4) % n
                else:
                    image, box = get_random_data(annotation_lines[i], input_shape)
                    i = (i+1) % n
                flag = bool(1-flag)
            else:
                image, box = get_random_data(annotation_lines[i], input_shape)
                i = (i+1) % n
            image_data.append(image)
            box_data.append(box)
        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)


#---------------------------------------------------#
#   读入xml文件，并输出y_true
#---------------------------------------------------#
def preprocess_true_boxes(true_boxes, input_shape, anchors, num_classes):
    assert (true_boxes[..., 4]<num_classes).all(), 'class id must be less than num_classes'
    # 一共有三个特征层数
    num_layers = len(anchors)//3
    # 先验框
    # 678为 142,110,  192,243,  459,401
    # 345为 36,75,  76,55,  72,146
    # 012为 12,16,  19,36,  40,28
    anchor_mask = [[6,7,8], [3,4,5], [0,1,2]] if num_layers==3 else [[3,4,5], [1,2,3]]

    true_boxes = np.array(true_boxes, dtype='float32')
    input_shape = np.array(input_shape, dtype='int32') # 416,416
    # 读出xy轴，读出长宽
    # 中心点(m,n,2)
    boxes_xy = (true_boxes[..., 0:2] + true_boxes[..., 2:4]) // 2
    boxes_wh = true_boxes[..., 2:4] - true_boxes[..., 0:2]
    # 计算比例
    true_boxes[..., 0:2] = boxes_xy/input_shape[:]
    true_boxes[..., 2:4] = boxes_wh/input_shape[:]

    # m张图
    m = true_boxes.shape[0]
    # 得到网格的shape为13,13;26,26;52,52
    grid_shapes = [input_shape//{0:32, 1:16, 2:8}[l] for l in range(num_layers)]
    # y_true的格式为(m,13,13,3,85)(m,26,26,3,85)(m,52,52,3,85)
    y_true = [np.zeros((m,grid_shapes[l][0],grid_shapes[l][1],len(anchor_mask[l]),5+num_classes),
        dtype='float32') for l in range(num_layers)]
    # [1,9,2]
    anchors = np.expand_dims(anchors, 0)
    anchor_maxes = anchors / 2.
    anchor_mins = -anchor_maxes
    # 长宽要大于0才有效
    valid_mask = boxes_wh[..., 0]>0

    for b in range(m):
        # 对每一张图进行处理
        wh = boxes_wh[b, valid_mask[b]]
        if len(wh)==0: continue
        # [n,1,2]
        wh = np.expand_dims(wh, -2)
        box_maxes = wh / 2.
        box_mins = -box_maxes

        # 计算真实框和哪个先验框最契合
        intersect_mins = np.maximum(box_mins, anchor_mins)
        intersect_maxes = np.minimum(box_maxes, anchor_maxes)
        intersect_wh = np.maximum(intersect_maxes - intersect_mins, 0.)
        intersect_area = intersect_wh[..., 0] * intersect_wh[..., 1]
        box_area = wh[..., 0] * wh[..., 1]
        anchor_area = anchors[..., 0] * anchors[..., 1]
        iou = intersect_area / (box_area + anchor_area - intersect_area)
        # 维度是(n) 感谢 消尽不死鸟 的提醒
        best_anchor = np.argmax(iou, axis=-1)

        for t, n in enumerate(best_anchor):
            for l in range(num_layers):
                if n in anchor_mask[l]:
                    # floor用于向下取整
                    i = np.floor(true_boxes[b,t,0]*grid_shapes[l][1]).astype('int32')
                    j = np.floor(true_boxes[b,t,1]*grid_shapes[l][0]).astype('int32')
                    # 找到真实框在特征层l中第b副图像对应的位置
                    k = anchor_mask[l].index(n)
                    c = true_boxes[b,t, 4].astype('int32')
                    y_true[l][b, j, i, k, 0:4] = true_boxes[b,t, 0:4]
                    y_true[l][b, j, i, k, 4] = 1
                    y_true[l][b, j, i, k, 5+c] = 1

    return y_true




if __name__ == "__main__":
    # 标签的位置
    annotation_path = '2007_train.txt'
    # 获取classes和anchor的位置
    classes_path = 'model_data/voc_classes.txt'    
    anchors_path = 'model_data/yolo_anchors.txt'
    # 预训练模型的位置
    weights_path = 'model_data/yolo4_weight.h5'
    # 获得classes和anchor
    class_names = get_classes(classes_path)
    anchors = get_anchors(anchors_path)
    # 一共有多少类
    num_classes = len(class_names)
    num_anchors = len(anchors)
    # 训练后的模型保存的位置
    log_dir = 'logs/'
    # 输入的shape大小
    # 显存比较小可以使用416x416
    # 现存比较大可以使用608x608
    input_shape = (416,416)
    mosaic = True
    Cosine_scheduler = False
    label_smoothing = 0

    # 清除session
    K.clear_session()

    # 输入的图像为
    image_input = Input(shape=(None, None, 3))
    h, w = input_shape

    # 创建yolo模型
    print('Create YOLOv3 model with {} anchors and {} classes.'.format(num_anchors, num_classes))
    model_body = yolo_body(image_input, num_anchors//3, num_classes)
    
    # 载入预训练权重
    print('Load weights {}.'.format(weights_path))
    model_body.load_weights(weights_path, by_name=True, skip_mismatch=True)
    
    # y_true为13,13,3,85
    # 26,26,3,85
    # 52,52,3,85
    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)]

    # 输入为*model_body.input, *y_true
    # 输出为model_loss
    loss_input = [*model_body.output, *y_true]
    model_loss = Lambda(yolo_loss, output_shape=(1,), name='yolo_loss',
        arguments={'anchors': anchors, 'num_classes': num_classes, 'ignore_thresh': 0.5, 'label_smoothing': label_smoothing})(loss_input)

    model = Model([model_body.input, *y_true], model_loss)

    # 训练参数设置
    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=False, period=1)
    early_stopping = EarlyStopping(monitor='val_loss', min_delta=0, patience=6, verbose=1)

    # 0.1用于验证，0.9用于训练
    val_split = 0.1
    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)
    num_train = len(lines) - num_val
    
    freeze_layers = 367
    for i in range(freeze_layers): model_body.layers[i].trainable = False
    print('Freeze the first {} layers of total {} layers.'.format(freeze_layers, len(model_body.layers)))

    # 调整非主干模型first
    if True:
        Init_epoch = 0
        Freeze_epoch = 25
        # batch_size大小，每次喂入多少数据
        batch_size = 8
        # 最大学习率
        learning_rate_base = 1e-3
        if Cosine_scheduler:
            # 预热期
            warmup_epoch = int((Freeze_epoch-Init_epoch)*0.2)
            # 总共的步长
            total_steps = int((Freeze_epoch-Init_epoch) * num_train / batch_size)
            # 预热步长
            warmup_steps = int(warmup_epoch * num_train / batch_size)
            # 学习率
            reduce_lr = WarmUpCosineDecayScheduler(learning_rate_base=learning_rate_base,
                                                        total_steps=(Freeze_epoch-Init_epoch),
                                                        warmup_learning_rate=1e-4,
                                                        warmup_steps=warmup_steps,
                                                        hold_base_rate_steps=num_train,
                                                        min_learn_rate=1e-6
                                                        )
            model.compile(optimizer=Adam(), loss={'yolo_loss': lambda y_true, y_pred: y_pred})
        else:
            reduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.5, patience=2, verbose=1)
            model.compile(optimizer=Adam(learning_rate_base), loss={'yolo_loss': lambda y_true, y_pred: y_pred})

        print('Train on {} samples, val on {} samples, with batch size {}.'.format(num_train, num_val, batch_size))
        model.fit_generator(data_generator(lines[:num_train], batch_size, input_shape, anchors, num_classes, mosaic=mosaic),
                steps_per_epoch=max(1, num_train//batch_size),
                validation_data=data_generator(lines[num_train:], batch_size, input_shape, anchors, num_classes, mosaic=mosaic),
                validation_steps=max(1, num_val//batch_size),
                epochs=Freeze_epoch,
                initial_epoch=Init_epoch,
                callbacks=[logging, checkpoint, reduce_lr, early_stopping])
        model.save_weights(log_dir + 'trained_weights_stage_1.h5')

    for i in range(freeze_layers): model_body.layers[i].trainable = True

    # 解冻后训练
    if True:
        Freeze_epoch = 25
        Epoch = 50
        # batch_size大小，每次喂入多少数据
        batch_size = 2

        # 最大学习率
        learning_rate_base = 1e-4
        if Cosine_scheduler:
            # 预热期
            warmup_epoch = int((Epoch-Freeze_epoch)*0.2)
            # 总共的步长
            total_steps = int((Epoch-Freeze_epoch) * num_train / batch_size)
            # 预热步长
            warmup_steps = int(warmup_epoch * num_train / batch_size)
            # 学习率
            reduce_lr = WarmUpCosineDecayScheduler(learning_rate_base=learning_rate_base,
                                                        total_steps=total_steps,
                                                        warmup_learning_rate=1e-5,
                                                        warmup_steps=warmup_steps,
                                                        hold_base_rate_steps=num_train//2,
                                                        min_learn_rate=1e-6
                                                        )
            model.compile(optimizer=Adam(), loss={'yolo_loss': lambda y_true, y_pred: y_pred})
        else:
            reduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.5, patience=2, verbose=1)
            model.compile(optimizer=Adam(learning_rate_base), loss={'yolo_loss': lambda y_true, y_pred: y_pred})

        print('Train on {} samples, val on {} samples, with batch size {}.'.format(num_train, num_val, batch_size))
        model.fit_generator(data_generator(lines[:num_train], batch_size, input_shape, anchors, num_classes, mosaic=mosaic),
                steps_per_epoch=max(1, num_train//batch_size//2.5) if mosaic else max(1, num_train//batch_size),
                validation_data=data_generator(lines[num_train:], batch_size, input_shape, anchors, num_classes, mosaic=mosaic),
                validation_steps=max(1, num_train//batch_size//2.5) if mosaic else max(1, num_train//batch_size),
                epochs=Epoch,
                initial_epoch=Freeze_epoch,
                callbacks=[logging, checkpoint, reduce_lr, early_stopping])
        model.save_weights(log_dir + 'last1.h5')