数据预处理：pandas的使用和cv2初步

最新推荐文章于 2024-05-02 01:06:36 发布

Rick@C137

最新推荐文章于 2024-05-02 01:06:36 发布

阅读量254

点赞数

分类专栏： Deep Learning

本文链接：https://blog.csdn.net/Aren8/article/details/115543568

版权

Deep Learning 专栏收录该内容

8 篇文章 0 订阅

订阅专栏

Pandas

解压

import zipfile
with zipfile.ZipFile("../input/train.zip","r") as z:
    z.extractall(".")
with zipfile.ZipFile("../input/test1.zip","r") as z:
    z.extractall(".")

创建dataframe

#读取文件夹里的所有文件名
filenames = os.listdir("../input/train/train")
#创建对应的分类
categories = []
for filename in filenames:
    category = filename.split('.')[0]
    if category == 'dog':
        categories.append(1)
    else:
        categories.append(0)
#创建dataframe
df = pd.DataFrame({
    'filename': filenames,
    'category': categories
})
#观察头尾
df.head()
df.tail()

观察标签

 df['category'].value_counts().plot.bar()

在这里插入图片描述

观察图片

sample = random.choice(filenames)
image = load_img("../input/train/train/"+sample)
plt.imshow(image)

train_test_split for dataframe
注意在split后trainning data和test data依然在同一个文件夹里

train_df, validate_df = train_test_split(df, test_size=0.20, random_state=42)
train_df = train_df.reset_index(drop=True)
validate_df = validate_df.reset_index(drop=True)
#drop=True，删除原本的index,使用新的index

dataframe大小

total_train = train_df.shape[0]
total_validate = validate_df.shape[0]

采样一行

example_df = train_df.sample(n=1).reset_index(drop=True)

submission

submission_df = test_df.copy()
submission_df['id'] = submission_df['filename'].str.split('.').str[0]
submission_df['label'] = submission_df['category']
submission_df.drop(['filename', 'category'], axis=1, inplace=True)
submission_df.to_csv('submission.csv', index=False)

Replace

test_df['category'] = test_df['category'].replace({ 'dog': 1, 'cat': 0 })

CV2

#使用lambda函数
convert = lambda category :(1 if category =='dog' else 0)
convert = lambda category : int(category == 'dog')

#读取图片及标签
category = convert(category)
img_array = cv2.imread(os.path.join(path,filename),
					cv2.IMREAD_GRAYSCALE) #灰度图片
new_img_array = cv2.resize(img_array, dsize=(80, 80))

X.append(new_img_array)
y.append(category)

#转numpy array，reshape,以及标准化
X = np.array(X).reshape(-1, 80,80,1)
y = np.array(y)
X/=255.0

保存数据集

import pickle
with open("train_x", "rb") as fp:
    X= pickle.load(fp)
with open("train_y", "rb") as fp:
    y= pickle.load(fp)
print(X.shape)

Transpose&Reshape

import numpy as np
arr_1d_bigger = np.arange(24)
arr_3d = arr_1d_bigger.reshape((2, 3, 4))
print(arr_3d)
arr = arr_3d.transpose(0,2,1)
print(arr)

[[[ 0  1  2  3]
  [ 4  5  6  7]
  [ 8  9 10 11]]

 [[12 13 14 15]
  [16 17 18 19]
  [20 21 22 23]]]
[[[ 0  4  8]
  [ 1  5  9]
  [ 2  6 10]
  [ 3  7 11]]