VerificationCode/dev/click_captcha/train_phrase.py

import os
import random
import sys
from glob import glob
os.environ["CUDA_VISIBLE_DEVICES"] = "0"
sys.path.append(os.path.dirname(os.path.abspath(__file__)))
sys.path.append(os.path.dirname(os.path.abspath(__file__)) + "/../")
from keras.callbacks import ReduceLROnPlateau, ModelCheckpoint
from keras.losses import BinaryCrossentropy, mse, CategoricalCrossentropy
from keras.metrics import Precision, Recall
from keras.optimizer_v2.adam import Adam
import keras.backend as K
from click_captcha.metrics import precision, recall, f1
from click_captcha.loss import focal_loss, contrastive_loss, l2_focal_loss, l1_focal_loss, l2_loss
from click_captcha.model import siamese_net, mobile_net, cnn_net, cnn_net_drag, u_net_drag, lstm_phrase, text_cnn_phrase
from click_captcha.pre_process import gen_mobile, gen_drag, gen_phrase

PRETRAINED = False
random.seed(42)
project_root = os.path.dirname(os.path.abspath(__file__)) + "/../"
vocabulary_len = 5792
sequence_len = 6


if __name__ == "__main__":
    model = text_cnn_phrase((sequence_len, vocabulary_len))

    if PRETRAINED:
        _path = "./models/e08-f10.86-phrase.h5"
        model.load_weights(_path, skip_mismatch=True, by_name=True)
        print("read pretrained model", _path)
    else:
        print("no pretrained")

    with open("../data/phrase/map2.txt", "r") as f:
        map_list = f.readlines()

    # data path split into train,test
    random.shuffle(map_list)
    testP = map_list[:int(len(map_list)*0.1)]
    trainP = map_list[int(len(map_list)*0.1):]
    print('total:', len(map_list), 'train:', len(trainP), 'test:', len(testP))

    # batch num
    batch_size = 64
    steps_per_epoch = max(1, len(trainP) // batch_size)
    validation_steps = max(1, len(testP) // batch_size)

    # 模型权重存放位置
    filepath = 'models/e{epoch:02d}-f1{val_f1:.2f}-phrase.h5'
    check_pointer = ModelCheckpoint(filepath=filepath, monitor='val_f1', verbose=0,
                                    save_weights_only=True, save_best_only=True,
                                    mode="max", save_freq='epoch')
    rlu = ReduceLROnPlateau(monitor='val_f1', factor=0.5, patience=10,
                            verbose=1, mode='max', cooldown=0, min_lr=0)
    model.compile(optimizer=Adam(lr=0.003), loss=BinaryCrossentropy(),
                  metrics=['acc', f1])

    # data loader
    train_loader = gen_phrase(trainP, batch_size=batch_size, shape=(vocabulary_len, sequence_len))
    test_loader = gen_phrase(testP, batch_size=batch_size, shape=(vocabulary_len, sequence_len))
    # train
    model.fit_generator(train_loader,
                        steps_per_epoch=steps_per_epoch,
                        callbacks=[check_pointer],
                        validation_data=test_loader,
                        validation_steps=validation_steps,
                        epochs=1000)