Skip to content

baseline

Author: https://www.kaggle.com/ivanwang2016

From: https://www.kaggle.com/ivanwang2016/baseline

License: Apache 2.0

In [1]:

# This Python 3 environment comes with many helpful analytics libraries installed
# It is defined by the kaggle/python docker image: https://github.com/kaggle/docker-python
# For example, here's several helpful packages to load in 

import numpy as np # linear algebra
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)

# Input data files are available in the "../input/" directory.
# For example, running this (by clicking run or pressing Shift+Enter) will list the files in the input directory

import os
print(os.listdir("../input"))
# Any results you write to the current directory are saved as output.


['test', 'sample_submission.csv', 'train.csv', 'train']

In [2]:

from PIL import Image
from tqdm import tqdm

from keras.preprocessing.image import ImageDataGenerator

def load_data(dataframe=None, batch_size=16, mode='categorical'):
    if dataframe is None:
        dataframe = pd.read_csv('../input/train.csv')
    dataframe['has_cactus'] = dataframe['has_cactus'].apply(str)
    gen = ImageDataGenerator(rescale=1./255., validation_split=0.1, horizontal_flip=True, vertical_flip=True)

    trainGen = gen.flow_from_dataframe(dataframe, directory='../input/train/train', x_col='id', y_col='has_cactus', has_ext=True, target_size=(32, 32),
        class_mode=mode, batch_size=batch_size, shuffle=True, subset='training')
    testGen = gen.flow_from_dataframe(dataframe, directory='../input/train/train', x_col='id', y_col='has_cactus', has_ext=True, target_size=(32, 32),
        class_mode=mode, batch_size=batch_size, shuffle=True, subset='validation')

    return trainGen, testGen


Using TensorFlow backend.

In [3]:

from keras.layers import Conv2D, MaxPool2D, Dense, BatchNormalization, Activation, GlobalAveragePooling2D
from keras.models import Sequential, Model
from keras.regularizers import l2

def baseline_model():
    model = Sequential()

    model.add(Conv2D(32, (3, 3), input_shape=(32, 32, 3), padding='same', use_bias=False, kernel_regularizer=l2(1e-4)))
    model.add(BatchNormalization())
    model.add(Activation('relu'))
    model.add(Conv2D(32, (3, 3), padding='same', use_bias=False, kernel_regularizer=l2(1e-4)))
    model.add(BatchNormalization())
    model.add(Activation('relu'))
    model.add(Conv2D(32, (3, 3), padding='same', use_bias=False, kernel_regularizer=l2(1e-4)))
    model.add(BatchNormalization())
    model.add(Activation('relu'))
    model.add(MaxPool2D())

    model.add(Conv2D(64, (3, 3), padding='same', use_bias=False, kernel_regularizer=l2(1e-4)))
    model.add(BatchNormalization())
    model.add(Activation('relu'))
    model.add(Conv2D(64, (3, 3), padding='same', use_bias=False, kernel_regularizer=l2(1e-4)))
    model.add(BatchNormalization())
    model.add(Activation('relu'))
    model.add(Conv2D(64, (3, 3), padding='same', use_bias=False, kernel_regularizer=l2(1e-4)))
    model.add(BatchNormalization())
    model.add(Activation('relu'))
    model.add(MaxPool2D())

    model.add(Conv2D(128, (3, 3), padding='same', use_bias=False, kernel_regularizer=l2(1e-4)))
    model.add(BatchNormalization())
    model.add(Activation('relu'))
    model.add(Conv2D(128, (3, 3), padding='same', use_bias=False, kernel_regularizer=l2(1e-4)))
    model.add(BatchNormalization())
    model.add(Activation('relu'))
    model.add(Conv2D(128, (3, 3), padding='same', use_bias=False, kernel_regularizer=l2(1e-4)))
    model.add(BatchNormalization())
    model.add(Activation('relu'))
    model.add(MaxPool2D())

    model.add(GlobalAveragePooling2D())
    model.add(Dense(2, activation='softmax'))

    return model

In [4]:

from keras.optimizers import Adam, SGD
from keras.callbacks import CSVLogger, ModelCheckpoint, ReduceLROnPlateau

def train_baseline():
    batch_size = 32
    trainGen, valGen = load_data(batch_size=batch_size)
    model = baseline_model()

    opt = Adam(1e-3)
    model.compile(loss='categorical_crossentropy', optimizer=opt, metrics=['accuracy'])
    cbs = [ReduceLROnPlateau(monitor='loss', factor=0.5, patience=1, min_lr=1e-5, verbose=1)]

    model.fit_generator(trainGen, steps_per_epoch=4922, epochs=3, validation_data=valGen, 
        validation_steps=493, shuffle=True, callbacks=cbs)

    return model

def predict_baseline(model):
    testdf = pd.read_csv('../input/sample_submission.csv')
    pred = np.empty((testdf.shape[0],))
    for n in tqdm(range(testdf.shape[0])):
        data = np.array(Image.open('../input/test/test/'+testdf.id[n]))
        data = data.astype(np.float32) / 255.
        pred[n] = model.predict(data.reshape((1, 32, 32, 3)))[0][1]

    testdf['has_cactus'] = pred
    testdf.to_csv('sample_submission.csv', index=False)

In [5]:

model = train_baseline()
predict_baseline(model)


Found 15750 images belonging to 2 classes.
Found 1750 images belonging to 2 classes.
Epoch 1/3
4922/4922 [==============================] - 247s 50ms/step - loss: 0.0674 - acc: 0.9889 - val_loss: 0.0786 - val_acc: 0.9813
Epoch 2/3
4922/4922 [==============================] - 230s 47ms/step - loss: 0.0326 - acc: 0.9965 - val_loss: 0.0428 - val_acc: 0.9888
Epoch 3/3
1400/4922 [=======>......................] - ETA: 2:35 - loss: 0.0273 - acc: 0.9974
回到顶部