Kaggle圖像分割比賽：keras平臺訓練unet++模型識別鹽沉積區（一）

Kaggle機器學習競賽是全球最著名的人工智能比賽，每一個競賽項目都吸引了大量AI愛好者參與。

這裏選擇2018年末進行的鹽沉積區識別競賽做爲例子：https://www.kaggle.com/c/tgs-salt-identification-challenge

1、數據

能夠從Kaggle網站下載，但需先註冊，下載速度可能也必將慢。能夠直接從個人百度網盤下載：

連接：https://pan.baidu.com/s/1htvnrwQagOXHXfjpaGedPQ
提取碼：a0zx

2、unet++模型開源代碼

unet++是2018年被提出的網絡模型，是對unet的優化，在圖像分割中有優異的表現。採用的源碼見：https://github.com/MrGiovanni/UNetPlusPlus

3、數據處理及準備

導入包：

import os
import random
import matplotlib.pyplot as plt

import pandas as pd
import numpy as np

from sklearn.model_selection import train_test_split
from skimage.transform import resize

from UNetPlusPlus_master.segmentation_models import Xnet

from keras.preprocessing.image import load_img
from keras.optimizers import *
from keras.callbacks import EarlyStopping, ModelCheckpoint

數據位置、尺寸：

root = r'E:\Kaggle\salt\competition_data'
model_path = root + '/model'

imgs_path = root + r'\train'
test_imgs_path = root + r'\test'

train_csv = root + r'\train.csv'
depths_csv = root + r'\depths.csv'
orig_img_w = 101
orig_img_h = 101

train_img_w = 224
train_img_h = 224

 原尺寸和訓練尺寸轉換：

def orig2tain(img):
    return resize(img, (train_img_w, train_img_h), mode='constant', preserve_range=True)
    
def train2orig(img):
    return resize(img, (orig_img_w, orig_img_h), mode='constant', preserve_range=True)

讀入數據：

train_df = pd.read_csv(train_csv, usecols=[0], index_col='id')
train_df["images"] = [np.array(load_img("{}/images/{}.png".format(imgs_path, idx), grayscale=False)) / 255 for idx in train_df.index]
train_df["masks"] = [np.array(load_img("{}/masks/{}.png".format(imgs_path, idx), grayscale=True)) / 255 for idx in train_df.index]

顯示讀入結果：

max_images = 10
grid_width = 10
grid_height = int(max_images / grid_width) + 1
fig, axs = plt.subplots(grid_height, grid_width, figsize=(20, 4))

for i, idx in enumerate(train_df.index[:max_images]):
    img = train_df.loc[idx].images
    mask = train_df.loc[idx].masks
    ax = axs[int(i / grid_width), i % grid_width]
    ax.imshow(img, cmap="Greys")
    ax = axs[int(i / grid_width)+1, i % grid_width]
    ax.imshow(mask, cmap="Greens")
    ax.set_yticklabels([])
    ax.set_xticklabels([])

plt.show()

 

 

 按2:8隨機分訓練集、驗證集：

train_ids, valid_ids, train_x, valid_x, train_y, valid_y = train_test_split(
    train_df.index.values,
    np.array(train_df.images.map(orig2tain).tolist()).reshape(-1, train_img_w, train_img_h, 3), 
    np.array(train_df.masks.map(orig2tain).tolist()).reshape(-1, train_img_w, train_img_h, 1), 
    test_size=0.2,
    random_state=123)

 

4、訓練

input_size = (train_img_w, train_img_h, 3)
model = Xnet(input_shape=input_size, backbone_name='resnet50', encoder_weights='imagenet', decoder_block_type='transpose')
model.compile(optimizer = Adam(lr = 1e-4), loss = 'binary_crossentropy', metrics = ['accuracy'])

model_name = 'Kaggle_Salt_{epoch:02d}-{val_acc:.3f}.hdf5'
abs_model_name = os.path.join(model_path, model_name)

model_checkpoint = ModelCheckpoint(abs_model_name, monitor='val_loss', verbose=2, save_best_only=True)
early_stop = EarlyStopping(monitor='val_loss', patience=6)
callbacks = [early_stop, model_checkpoint]

history = model.fit(train_x, train_y,
                    validation_data=[valid_x, valid_y], 
                    epochs=100,
                    batch_size=4,
                    callbacks=callbacks)

 

顯示訓練曲線：

    acc = history.history['acc']
    val_acc = history.history['val_acc']
    loss = history.history['loss']
    val_loss = history.history['val_loss']

    epochs = range(len(acc))

    plt.plot(epochs, acc, 'bo', label='Training acc')
    plt.plot(epochs, val_acc, 'b', label='Validation acc')
    plt.title('Training and validation accuracy')
    plt.legend()
                
    plt.figure()

    plt.plot(epochs, loss, 'bo', label='Training loss')
    plt.plot(epochs, val_loss, 'b', label='Validation loss')
    plt.title('Training and validation loss')
    plt.legend()
    
    plt.show()