# -*- coding: utf-8 -*- import osimport numpy as npimport pandas as pdimport h5pyimport pylabimport matplotlib.pyplot as plt trainpath = str('C:/Users/49691/Desktop/数据集/train/')testpath = str('C:/Users/49691/Desktop/数据集/test/')n_tr = len(os.listdir(trainpath))print('num of training files: ', n_tr) train_labels = pd.read_csv('C:/Users/49691/Desktop/数据集/sample_submission.csv')train_labels.head() from skimage import io, transform x = np.empty(shape=(n_tr, 224, 224, 3))y = np.empty(n_tr) labels = train_labels.invasive.valuesname = train_labels.name.values permutation=np.random.permutation(name.shape[0])print(permutation)print(labels[permutation])save_data = pd.DataFrame({'name':permutation,'invasive':labels[permutation]})save_data.to_csv('C:/Users/49691/Desktop/数据集/b.csv') for k,v in enumerate(np.random.permutation(n_tr)): print(k,v) path = '{0}{1}.jpg'.format(trainpath, v) tr_im = io.imread(path) x[k] = transform.resize(tr_im, output_shape=(224, 224, 3)) y[k] = float(labels[v-1]) 
