hi, how can i use all classes rather than the classes with over 400 pixel?

[zhshining]

hi, i was trying to use 1D_CNN to classcify all 16 classes, but when I rectified the 'data_loader' and 'get_result', the program reported wrong:'ValueError: Sample larger than population or is negative'. My revisions were as follow: How can I do? please kindly tell me.

[zhangjinyangnwpu]

I just use 9 classes in Indian Pines dataset due to the sample unbalance，if you want to use all categories, just ignore the code

      # comment these code
      if self.data_name == 'Indian_pines':
            imGIS = data_gt
            origin_num = np.zeros(shape=[17], dtype=int)
            for i in range(imGIS.shape[0]):
                for j in range(imGIS.shape[1]):
                    for k in range(1, 17):
                        if imGIS[i][j] == k:
                            origin_num[k] += 1
            index = 0
            data_num = np.zeros(shape=[9], dtype=int)  # per calsses's num
            data_label = np.zeros(shape=[9], dtype=int)  # original labels
            for i in range(len(origin_num)):
                if origin_num[i] > 400:
                    data_num[index] = origin_num[i]
                    data_label[index] = i
                    index += 1
            iG = np.zeros([imGIS.shape[0], imGIS.shape[1]], dtype=imGIS.dtype)
            for i in range(imGIS.shape[0]):
                for j in range(imGIS.shape[1]):
                    if imGIS[i, j] in data_label:
                        for k in range(len(data_label)):
                            if imGIS[i][j] == data_label[k]:
                                iG[i, j] = k + 1
                                continue
            imGIS = iG

            data_gt = imGIS
            self.data_gt = data_gt

in data_loader.py， and set your own train number for peer class in

for k, v in data_pos.items():
     if self.train_num > 0 and self.train_num < 1:
           train_num = self.train_num * len(v)# set your percentage
     else:
           train_num = self.train_num# set your own train per class, notice some class may don't have sampel more than 400 in IndianPines
      train_pos[k] = random.sample(v, int(train_num))
      test_pos[k] = [i for i in v if i not in train_pos[k]]