split_data.py

import numpy as np
from sklearn.decomposition import PCA
def split_data(gt_reshape, class_num, train_ratio, val_ratio, train_num, val_num, samples_type):
    train_index = []
    test_index = []
    val_index = []
    if samples_type == 'ratio':
        # class_num = 16 类
        for i in range(class_num):

            idx = np.where(gt_reshape == i + 1)[-1] 
            samplesCount = len(idx)
            # print("Class ",i,":", samplesCount)  
            train_num = np.ceil(samplesCount * train_ratio).astype('int32')  
            val_num = np.ceil(samplesCount * val_ratio).astype('int32')  
            np.random.shuffle(idx)
            train_index.append(idx[:train_num])
            val_index.append(idx[train_num:train_num+val_num])
            test_index.append(idx[train_num+val_num:])

    else:
        sample_num = train_num
        # class_num = 16 类
        for i in range(class_num):
            idx = np.where(gt_reshape == i + 1)[-1] 
            samplesCount = len(idx)
            # print("Class ",i,":", samplesCount)  # 每一类的个数

            max_index = np.max(samplesCount) + 1
            np.random.shuffle(idx)
            if sample_num > max_index:
                sample_num = 10
            else:
                sample_num = train_num

            # 取出每个类别选择出的训练集
            train_index.append(idx[: sample_num])
            val_index.append(idx[sample_num : sample_num+val_num])
            test_index.append(idx[sample_num+val_num : ])

    train_index = np.concatenate(train_index, axis=0)
    val_index = np.concatenate(val_index, axis=0)
    test_index = np.concatenate(test_index, axis=0)

    return train_index, val_index, test_index

def apply_PCA(X, num_components=75):
    newX = np.reshape(X, (-1, X.shape[2]))
    pca = PCA(n_components=num_components, whiten=True)
    newX = pca.fit_transform(newX)
    newX = np.reshape(newX, (X.shape[0],X.shape[1], num_components))
    return newX, pca