machine-learning - 我从 kaggle 下载了一个 DL 模型，该模型用于 keras 的 8 分类器工作

Question

我从kaggle下载了一个DL模型，该模型用于keras的8分类器工作。显然，它在模型顶部有密集(8)。我想将它用于2分类器，所以我将顶部完整连接从密集(8)修改为密集(2)。我认为这对我有用。但是，当我运行脚本时，终端报告错误。感谢您的耐心和帮助。错误如下

Error when checking model target: expected dense_3 to have shape (None, 2) but got array with shape (1333L, 8L) 

这是代码，可能很长

# %load kaggle_dog_cat_classifier.py
__author__ = 'JeofuHuang: https://www.kaggle.com/jeofuhuang'
import numpy as np
np.random.seed(2016)

import os
import glob
import cv2
import datetime
import pandas as pd
import time
import warnings
warnings.filterwarnings("ignore")

from sklearn.cross_validation import KFold
from keras.models import Sequential
from keras.layers.core import Dense, Dropout, Flatten
from keras.layers.convolutional import Convolution2D, MaxPooling2D, ZeroPadding2D
from keras.optimizers import SGD
from keras.callbacks import EarlyStopping
from keras.utils import np_utils
from sklearn.metrics import log_loss
from keras import __version__ as keras_version

def get_im_cv2(path):
    img = cv2.imread(path)
    resized = cv2.resize(img, (32, 32), cv2.INTER_LINEAR)
    return resized


def load_train():
    X_train = []
    X_train_id = []
    y_train = []
    start_time = time.time()

    print('Read train images')
    folders = ['dog', 'cat']
    for fld in folders:
        index = folders.index(fld)
        print('Load folder {} (Index: {})'.format(fld, index))
        path = os.path.join('.', 'input', 'train', fld, '*.jpg')
        files = glob.glob(path)
        for fl in files:
            flbase = os.path.basename(fl)
            img = get_im_cv2(fl)
            X_train.append(img)
            X_train_id.append(flbase)
            y_train.append(index)

    print('Read train data time: {} seconds'.format(round(time.time() - start_time, 2)))
    return X_train, y_train, X_train_id


def load_test():
    path = os.path.join('.', 'input', 'test', '*.jpg')
    files = sorted(glob.glob(path))

    X_test = []
    X_test_id = []
    for fl in files:
        flbase = os.path.basename(fl)
        img = get_im_cv2(fl)
        X_test.append(img)
        X_test_id.append(flbase)

    return X_test, X_test_id


def create_submission(predictions, test_id, info):
    result1 = pd.DataFrame(predictions, columns=['dog', 'cat'])
    result1.loc[:, 'image'] = pd.Series(test_id, index=result1.index)
    now = datetime.datetime.now()
    sub_file = 'submission_' + info + '_' + str(now.strftime("%Y-%m-%d-%H-%M")) + '.csv'
    result1.to_csv(sub_file, index=False)


def read_and_normalize_train_data():
    train_data, train_target, train_id = load_train()

    print('Convert to numpy...')
    train_data = np.array(train_data, dtype=np.uint8)
    train_target = np.array(train_target, dtype=np.uint8)

    print('Reshape...')
    train_data = train_data.transpose((0, 3, 1, 2))

    print('Convert to float...')
    train_data = train_data.astype('float32')
    train_data = train_data / 255
    train_target = np_utils.to_categorical(train_target, 8)

    print('Train shape:', train_data.shape)
    print(train_data.shape[0], 'train samples')
    return train_data, train_target, train_id


def read_and_normalize_test_data():
    start_time = time.time()
    test_data, test_id = load_test()

    test_data = np.array(test_data, dtype=np.uint8)
    test_data = test_data.transpose((0, 3, 1, 2))

    test_data = test_data.astype('float32')
    test_data = test_data / 255

    print('Test shape:', test_data.shape)
    print(test_data.shape[0], 'test samples')
    print('Read and process test data time: {} seconds'.format(round(time.time() - start_time, 2)))
    return test_data, test_id


def dict_to_list(d):
    ret = []
    for i in d.items():
        ret.append(i[1])
    return ret


def merge_several_folds_mean(data, nfolds):
    a = np.array(data[0])
    for i in range(1, nfolds):
        a += np.array(data[i])
    a /= nfolds
    return a.tolist()


def create_model():
    model = Sequential()
    model.add(ZeroPadding2D((1, 1), input_shape=(3, 32, 32), dim_ordering='th'))
    model.add(Convolution2D(4, 3, 3, activation='relu', dim_ordering='th'))
    model.add(ZeroPadding2D((1, 1), dim_ordering='th'))
    model.add(Convolution2D(4, 3, 3, activation='relu', dim_ordering='th'))
    model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2), dim_ordering='th'))

    model.add(ZeroPadding2D((1, 1), dim_ordering='th'))
    model.add(Convolution2D(8, 3, 3, activation='relu', dim_ordering='th'))
    model.add(ZeroPadding2D((1, 1), dim_ordering='th'))
    model.add(Convolution2D(8, 3, 3, activation='relu', dim_ordering='th'))
    model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2), dim_ordering='th'))

    model.add(Flatten())
    model.add(Dense(32, activation='relu'))
    model.add(Dropout(0.5))
    model.add(Dense(32, activation='relu'))
    model.add(Dropout(0.5))
    model.add(Dense(2, activation='softmax'))

    sgd = SGD(lr=1e-2, decay=1e-6, momentum=0.9, nesterov=True)
    model.compile(optimizer=sgd, loss='categorical_crossentropy')

    return model


def get_validation_predictions(train_data, predictions_valid):
    pv = []
    for i in range(len(train_data)):
        pv.append(predictions_valid[i])
    return pv


def run_cross_validation_create_models(nfolds=10):
    # input image dimensions
    batch_size = 16
    nb_epoch = 30
    random_state = 51

    train_data, train_target, train_id = read_and_normalize_train_data()

    yfull_train = dict()
    kf = KFold(len(train_id), n_folds=nfolds, shuffle=True, 
random_state=random_state)
    num_fold = 0
    sum_score = 0
    models = []
    for train_index, test_index in kf:
        model = create_model()
        X_train = train_data[train_index]
        Y_train = train_target[train_index]
        X_valid = train_data[test_index]
        Y_valid = train_target[test_index]

        num_fold += 1
        print('Start KFold number {} from {}'.format(num_fold, nfolds))
        print('Split train: ', len(X_train), len(Y_train))
        print('Split valid: ', len(X_valid), len(Y_valid))

        callbacks = [
            EarlyStopping(monitor='val_loss', patience=3, verbose=0),
        ]
        model.fit(X_train, Y_train, batch_size=batch_size, 
            nb_epoch=nb_epoch,shuffle=True, verbose=2, validation_data=
            (X_valid,  Y_valid), callbacks=callbacks)

        predictions_valid = model.predict(X_valid.astype('float32'), 
            batch_size=batch_size, verbose=2)
        score = log_loss(Y_valid, predictions_valid)
        print('Score log_loss: ', score)
        sum_score += score*len(test_index)

        # Store valid predictions
        for i in range(len(test_index)):
            yfull_train[test_index[i]] = predictions_valid[i]

        models.append(model)

    score = sum_score/len(train_data)
    print("Log_loss train independent avg: ", score)

    info_string = 'loss_' + str(score) + '_folds_' + str(nfolds) + '_ep_' + 
str(nb_epoch)
    return info_string, models


def run_cross_validation_process_test(info_string, models):
    batch_size = 16
    num_fold = 0
    yfull_test = []
    test_id = []
    nfolds = len(models)

    for i in range(nfolds):
        model = models[i]
        num_fold += 1
        print('Start KFold number {} from {}'.format(num_fold, nfolds))
        test_data, test_id = read_and_normalize_test_data()
        test_prediction = model.predict(test_data, batch_size=batch_size, 
verbose=2)
        yfull_test.append(test_prediction)

    test_res = merge_several_folds_mean(yfull_test, nfolds)
    info_string = 'loss_' + info_string \ + '_folds_' + str(nfolds)
    create_submission(test_res, test_id, info_string)


if __name__ == '__main__':
    print('Keras version: {}'.format(keras_version))
    num_folds = 3
    info_string, models = run_cross_validation_create_models(num_folds)
    run_cross_validation_process_test(info_string, models)

Answer 1

错误出现在read_and_normalize_train_data()函数中修改

train_target = np_utils.to_categorical(train_target, 8)

至

train_target = np_utils.to_categorical(train_target, 2)

然后就可以了。