xinge008 · yexinyinancy · Jun 14, 2021 · Jun 14, 2021 · Jun 23, 2021 · Jun 23, 2021
diff --git a/.gitignore b/.gitignore
diff --git a/NUSCENES-GUIDE.md b/NUSCENES-GUIDE.md
diff --git a/README.md b/README.md
diff --git a/builder/__init__.py b/builder/__init__.py
diff --git a/builder/__pycache__/__init__.cpython-37.pyc b/builder/__pycache__/__init__.cpython-37.pyc
diff --git a/builder/__pycache__/data_builder.cpython-37.pyc b/builder/__pycache__/data_builder.cpython-37.pyc
diff --git a/builder/__pycache__/loss_builder.cpython-37.pyc b/builder/__pycache__/loss_builder.cpython-37.pyc
diff --git a/builder/__pycache__/model_builder.cpython-37.pyc b/builder/__pycache__/model_builder.cpython-37.pyc
diff --git a/builder/data_builder.py b/builder/data_builder.py
@@ -30,7 +30,8 @@ def build(dataset_config,
                                 return_ref=train_ref, label_mapping=label_mapping, nusc=nusc)
     val_pt_dataset = SemKITTI(data_path, imageset=val_imageset,
                               return_ref=val_ref, label_mapping=label_mapping, nusc=nusc)
-
+    #import pdb
+    #pdb.set_trace()
     train_dataset = get_model_class(dataset_config['dataset_type'])(
         train_pt_dataset,
         grid_size=grid_size,

diff --git a/builder/loss_builder.py b/builder/loss_builder.py
diff --git a/builder/model_builder.py b/builder/model_builder.py
diff --git a/config/__init__.py b/config/__init__.py
diff --git a/config/__pycache__/__init__.cpython-37.pyc b/config/__pycache__/__init__.cpython-37.pyc
diff --git a/config/__pycache__/config.cpython-37.pyc b/config/__pycache__/config.cpython-37.pyc
diff --git a/config/config.py b/config/config.py
diff --git a/config/label_mapping/.semantic-kitti.yaml.swp b/config/label_mapping/.semantic-kitti.yaml.swp
diff --git a/config/label_mapping/nuscenes.yaml b/config/label_mapping/nuscenes.yaml
diff --git a/config/label_mapping/semantic-kitti-multiscan.yaml b/config/label_mapping/semantic-kitti-multiscan.yaml
diff --git a/config/label_mapping/semantic-kitti.yaml b/config/label_mapping/semantic-kitti.yaml
@@ -185,18 +185,19 @@ learning_ignore: # Ignore classes
   19: False    # "traffic-sign"
 split: # sequence numbers
   train:
-    - 0
-    - 1
-    - 2
+#    - 0
+#    - 1
+#    - 2
     - 3
     - 4
-    - 5
-    - 6
-    - 7
+#    - 5
+#    - 6
+#    - 7
     - 9
     - 10
   valid:
-    - 8
+#    - 8
+    - 22
   test:
     - 11
     - 12

diff --git a/config/nuScenes.yaml b/config/nuScenes.yaml
diff --git a/config/semantickitti.yaml b/config/semantickitti.yaml
@@ -41,15 +41,15 @@ dataset_params:
 ###################
 ## Data_loader options
 train_data_loader:
-  data_path: "/data/dataset/semantic_kitti/data_semkitti/dataset/sequences/"
+  data_path: "/lustre/home/acct-stu/stu010/sequences/"
   imageset: "train"
   return_ref: True
   batch_size: 2
   shuffle: True
   num_workers: 4
 
 val_data_loader:
-  data_path: "/data/dataset/semantic_kitti/data_semkitti/dataset/sequences/"
+  data_path: "/lustre/home/acct-stu/stu010/sequences/"
   imageset: "val"
   return_ref: True
   batch_size: 1
@@ -62,7 +62,7 @@ val_data_loader:
 train_params:
   model_load_path: "./model_load_dir/model_load.pt"
   model_save_path: "./model_save_dir/model_save.pt"
-  checkpoint_every_n_steps: 4599
+  checkpoint_every_n_steps: 1932
   max_num_epochs: 40
-  eval_every_n_steps: 4599
+  eval_every_n_steps: 1932
   learning_rate: 0.001
diff --git a/dataloader/__init__.py b/dataloader/__init__.py
diff --git a/dataloader/__pycache__/__init__.cpython-37.pyc b/dataloader/__pycache__/__init__.cpython-37.pyc
diff --git a/dataloader/__pycache__/dataset_nuscenes.cpython-37.pyc b/dataloader/__pycache__/dataset_nuscenes.cpython-37.pyc
diff --git a/dataloader/__pycache__/dataset_nuscenes.nb_process_label-137.py37m.1.nbc b/dataloader/__pycache__/dataset_nuscenes.nb_process_label-137.py37m.1.nbc
diff --git a/dataloader/__pycache__/dataset_nuscenes.nb_process_label-137.py37m.2.nbc b/dataloader/__pycache__/dataset_nuscenes.nb_process_label-137.py37m.2.nbc
diff --git a/dataloader/__pycache__/dataset_semantickitti.nb_process_label-379.py37m.1.nbc b/dataloader/__pycache__/dataset_semantickitti.nb_process_label-379.py37m.1.nbc
diff --git a/dataloader/__pycache__/dataset_semantickitti.nb_process_label-379.py37m.2.nbc b/dataloader/__pycache__/dataset_semantickitti.nb_process_label-379.py37m.2.nbc
diff --git a/dataloader/__pycache__/dataset_semantickitti.nb_process_label-391.py37m.1.nbc b/dataloader/__pycache__/dataset_semantickitti.nb_process_label-391.py37m.1.nbc
diff --git a/dataloader/__pycache__/dataset_semantickitti.nb_process_label-391.py37m.nbi b/dataloader/__pycache__/dataset_semantickitti.nb_process_label-391.py37m.nbi
diff --git a/dataloader/__pycache__/dataset_semantickitti.nb_process_label-392.py37m.1.nbc b/dataloader/__pycache__/dataset_semantickitti.nb_process_label-392.py37m.1.nbc
diff --git a/dataloader/__pycache__/dataset_semantickitti.nb_process_label-392.py37m.nbi b/dataloader/__pycache__/dataset_semantickitti.nb_process_label-392.py37m.nbi
diff --git a/dataloader/__pycache__/dataset_semantickitti.nb_process_label-405.py37m.1.nbc b/dataloader/__pycache__/dataset_semantickitti.nb_process_label-405.py37m.1.nbc
diff --git a/dataloader/__pycache__/dataset_semantickitti.nb_process_label-405.py37m.nbi b/dataloader/__pycache__/dataset_semantickitti.nb_process_label-405.py37m.nbi
diff --git a/dataloader/dataset_nuscenes.py b/dataloader/dataset_nuscenes.py
diff --git a/dataloader/dataset_semantickitti.py b/dataloader/dataset_semantickitti.py
@@ -16,6 +16,13 @@
 
 REGISTERED_DATASET_CLASSES = {}
 
+# triple
+def triple_func(x):
+    return x ** (1/3)
+
+# ln
+def ln_func(x):
+    return np.log(x)
 
 def register_dataset(cls, name=None):
     global REGISTERED_DATASET_CLASSES
@@ -244,7 +251,26 @@ def __getitem__(self, index):
         intervals = crop_range / (cur_grid_size - 1)
 
         if (intervals == 0).any(): print("Zero interval!")
-        grid_ind = (np.floor((np.clip(xyz_pol, min_bound, max_bound) - min_bound) / intervals)).astype(np.int)
+        # fixed
+        # grid_ind = (np.floor((np.clip(xyz_pol, min_bound, max_bound) - min_bound) / intervals)).astype(np.int)
+
+        # triple (quadratic interval)
+        #x_clip = np.clip(xyz_pol, min_bound, max_bound)
+        #a = (cur_grid_size[0] - 1) / (triple_func(max_bound[0] - min_bound[0]))
+        #tmp_y = a * triple_func(x_clip[:, 0] - min_bound[0])
+        #grid_ind1 = (np.floor(tmp_y)).astype(np.int)
+        #grid_ind2 = (np.floor((x_clip[:, 1:] - min_bound[1:]) / ((max_bound[1:] - min_bound[1:]) / (cur_grid_size[1:] - 1)))).astype(np.int)
+        #grid_ind = np.concatenate((grid_ind1.reshape((grid_ind1.shape[0], 1)), grid_ind2), axis=1)
+        ##################
+
+        #exponential (exponential interval)
+        x_clip = np.clip(xyz_pol, min_bound, max_bound)
+        a = (cur_grid_size[0] - 1) / (ln_func(max_bound[0] + 1 - min_bound[0]))
+        tmp_y = a * ln_func(x_clip[:, 0] + 1 - min_bound[0])
+        grid_ind1 = (np.floor(tmp_y)).astype(np.int)
+        grid_ind2 = (np.floor((x_clip[:, 1:] - min_bound[1:]) / ((max_bound[1:] - min_bound[1:]) / (cur_grid_size[1:] - 1)))).astype(np.int)
+        grid_ind = np.concatenate((grid_ind1.reshape((grid_ind1.shape[0], 1)), grid_ind2), axis=1)
+        ######################
 
         voxel_position = np.zeros(self.grid_size, dtype=np.float32)
         dim_array = np.ones(len(self.grid_size) + 1, int)

diff --git a/dataloader/pc_dataset.py b/dataloader/pc_dataset.py
diff --git a/demo_folder.py b/demo_folder.py
diff --git a/img/leaderboard.png b/img/leaderboard.png
diff --git a/img/leaderboard2.png b/img/leaderboard2.png
diff --git a/img/pipeline.png b/img/pipeline.png
diff --git a/logs_dir/cylinder_asym_networks_logs_tee.txt b/logs_dir/cylinder_asym_networks_logs_tee.txt
diff --git a/network/__init__.py b/network/__init__.py
diff --git a/network/__pycache__/__init__.cpython-37.pyc b/network/__pycache__/__init__.cpython-37.pyc
diff --git a/network/__pycache__/cylinder_fea_generator.cpython-37.pyc b/network/__pycache__/cylinder_fea_generator.cpython-37.pyc
diff --git a/network/__pycache__/cylinder_spconv_3d.cpython-37.pyc b/network/__pycache__/cylinder_spconv_3d.cpython-37.pyc
diff --git a/network/__pycache__/segmentator_3d_asymm_spconv.cpython-37.pyc b/network/__pycache__/segmentator_3d_asymm_spconv.cpython-37.pyc
diff --git a/network/cylinder_fea_generator.py b/network/cylinder_fea_generator.py
diff --git a/network/cylinder_spconv_3d.py b/network/cylinder_spconv_3d.py
diff --git a/network/segmentator_3d_asymm_spconv.py b/network/segmentator_3d_asymm_spconv.py
@@ -46,6 +46,13 @@ def conv1x1(in_planes, out_planes, stride=1, indice_key=None):
 class ResContextBlock(nn.Module):
     def __init__(self, in_filters, out_filters, kernel_size=(3, 3, 3), stride=1, indice_key=None):
         super(ResContextBlock, self).__init__()
+        ####no asymconv
+        # self.conv1 = conv3x3(in_filters, out_filters, indice_key=indice_key + "bef")
+        # self.bn0 = nn.BatchNorm1d(out_filters)
+        # self.act1 = nn.LeakyReLU()
+        # self.weight_initialization()
+        ##################
+
         self.conv1 = conv1x3(in_filters, out_filters, indice_key=indice_key + "bef")
         self.bn0 = nn.BatchNorm1d(out_filters)
         self.act1 = nn.LeakyReLU()
@@ -71,6 +78,13 @@ def weight_initialization(self):
                 nn.init.constant_(m.bias, 0)
 
     def forward(self, x):
+        ####noasymconv
+        # shortcut = self.conv1(x)
+        # shortcut.features = self.act1(shortcut.features)
+        # shortcut.features = self.bn0(shortcut.features)
+        # return shortcut
+        ############
+
         shortcut = self.conv1(x)
         shortcut.features = self.act1(shortcut.features)
         shortcut.features = self.bn0(shortcut.features)

diff --git a/statistic_class_num.py b/statistic_class_num.py
@@ -0,0 +1,82 @@
+import numpy as np
+import os
+import tqdm
+import matplotlib.mlab as mlab
+import matplotlib.pyplot as plt
+from sklearn.model_selection import train_test_split
+import random
+import shutil
+
+def plot():
+    class_map = {5: 'bus', 4: 'truck', 8: 'motorcyclist', 6: 'person', 7: 'bicyclist', 1: 'car', 0: 'unlabeled',
+                 19: 'traffic-sign',18: 'pole', 17: 'terrain', 16: 'trunk', 15: 'vegetation', 9: 'road', 14: 'fence', 13: 'building',
+                 12: 'other-ground', 11: 'sidewalk', 10: 'parking', 3: 'motorcycle', 2: 'bicycle'}
+    file_num = 11
+    for j in range(file_num):
+        X = [str(i) for i in range(20)]
+        Y = []
+        fig = plt.figure()
+        with open('out' + str(j).zfill(2) + '.txt', 'r') as f:
+            for line in f.readlines():
+                line = line.strip('\n')  # 去掉列表中每一个元素的换行符
+                Y.append(int(line))
+        plt.bar(X, Y, 0.4, color="blue")
+        plt.xlabel("X-axis")
+        plt.ylabel("Y-axis")
+        plt.title('Sequence' + str(j).zfill(2))
+
+        # plt.show()
+        plt.savefig('barChart' + str(j).zfill(2) + '.jpg')
+
+
+def statistc_class():
+    learning_map = {0: 0, 1: 0, 10: 1, 11: 2, 13: 5, 15: 3, 16: 5, 18: 4, 20: 5, 30: 6, 31: 7, 32: 8, 40: 9, 44: 10,
+                    48: 11, 49: 12, 50: 13, 51: 14, 52: 0, 60: 9, 70: 15, 71: 16, 72: 17, 80: 18, 81: 19, 99: 0, 252: 1,
+                    253: 7, 254: 6, 255: 8, 256: 5, 257: 5, 258: 4, 259: 5}
+
+    statistc_class_dic = {}
+    file_num = 11
+    root_path = '/Users/yexinyi/Desktop/VE450.nosync/data/merge/dataset/sequences'
+    for i in range(20):
+        statistc_class_dic[i] = 0
+
+    for j in range(file_num):
+        folder_path = os.path.join(root_path, str(j).zfill(2), 'labels')
+        # folder_path = '/Users/yexinyi/Desktop/VE450.nosync/data/val_sub/dataset/sequences/08/labels'
+        for root, dirs, files in os.walk(folder_path):
+            for f in tqdm.tqdm(files):
+                # filepath='/Users/yexinyi/Desktop/VE450.nosync/data/merge/dataset/sequences/00/labels/003907.label'
+                filepath = os.path.join(root, f)
+                annotated_data = np.fromfile(filepath, dtype=np.uint32).reshape((-1, 1))
+                annotated_data = annotated_data & 0xFFFF  # delete high 16 digits binary
+                tmp = annotated_data.astype(np.uint8)
+                for i in range(len(tmp)):
+                    if tmp[i][0] in learning_map.keys():
+                        if learning_map[tmp[i][0]] in statistc_class_dic.keys():
+                            statistc_class_dic[learning_map[tmp[i][0]]] += 1
+
+        with open('out_val' + str(j).zfill(2) + '.txt', 'w') as file:
+            for i in statistc_class_dic.values():
+                file.write(str(i) + '\n')
+
+        break
+
+def split_dataset():
+    folder_path = '/Users/yexinyi/Desktop/VE450.nosync/data/merge/dataset/sequences/08/labels'
+    folder_path2 = '/Users/yexinyi/Desktop/VE450.nosync/data/merge/dataset/sequences/08/velodyne'
+
+    file_list = random.sample(os.listdir(folder_path), 814)  #0.2*4070
+    for i in tqdm.tqdm(file_list):
+        source = os.path.join(folder_path, i)
+        target = os.path.join('/Users/yexinyi/Desktop/VE450.nosync/data/val_sub/dataset/sequences/08/labels', i)
+        source2 = os.path.join(folder_path2, i[:-5]+'bin')
+        target2 = os.path.join('/Users/yexinyi/Desktop/VE450.nosync/data/val_sub/dataset/sequences/08/velodyne', i[:-5]+'bin')
+        shutil.copy(source, target)
+        shutil.copy(source2, target2)
+
+
+
+if __name__ == '__main__':
+    statistc_class()
+    # plot()
+    # split_dataset()
diff --git a/statistic_new_data_coding.py b/statistic_new_data_coding.py
@@ -0,0 +1,166 @@
+import numpy as np
+import os
+import tqdm
+import matplotlib.mlab as mlab
+import matplotlib.pyplot as plt
+from sklearn.model_selection import train_test_split
+import random
+import shutil
+
+# transformation between Cartesian coordinates and polar coordinates
+def cart2polar(input_xyz):
+    rho = np.sqrt(input_xyz[:, 0] ** 2 + input_xyz[:, 1] ** 2)
+    phi = np.arctan2(input_xyz[:, 1], input_xyz[:, 0])
+    return np.stack((rho, phi, input_xyz[:, 2]), axis=1)
+
+# quadratic
+def quadratic_func(x):
+    return x ** 0.5
+
+# triple
+def triple_func(x):
+    return x ** (1/3)
+
+# ln
+def ln_func(x):
+    return np.log(x)
+
+def cal_cell_num(func_type):
+    file_num = 11
+    root_path = '/Users/yexinyi/Desktop/VE450.nosync/data/merge/dataset/sequences'
+
+    max_bound = np.asarray([50, 3.1415926, 2])
+    min_bound = np.asarray([0, -3.1415926, -4])
+    crop_range = max_bound - min_bound
+    cur_grid_size = np.asarray([480, 360, 32])
+
+    # for plot
+    dim1 = cur_grid_size[0]
+    dim2 = cur_grid_size[1]
+    dim3 = cur_grid_size[2]
+
+    total_cell_list = [0, 0, 0, 0, 0]
+    full_cell_list = [0, 0, 0, 0, 0]
+    res = []
+    total_file_num = 0
+
+    for j in range(file_num):
+        if j == 8:
+            continue
+        folder_path = os.path.join(root_path, str(j).zfill(2), 'velodyne')
+        # folder_path = '/Users/yexinyi/Desktop/VE450.nosync/data/merge/dataset/sequences/01/velodyne'
+        for root, dirs, files in os.walk(folder_path):
+            for f in tqdm.tqdm(files):
+                # filepath='/Users/yexinyi/Desktop/VE450.nosync/data/merge/dataset/sequences/00/velodyne/003907.bin'
+                filepath = os.path.join(root, f)
+                raw_data = np.fromfile(filepath, dtype=np.float32).reshape((-1, 4))
+                xyz_pol = cart2polar(raw_data[:, :3])
+
+                total_file_num += 1
+                if func_type == "quadratic":
+                    x_clip = np.clip(xyz_pol, min_bound, max_bound)
+                    a = (cur_grid_size[0] - 1) / (quadratic_func(max_bound[0] - min_bound[0]))
+                    tmp_y = a * quadratic_func(x_clip[:, 0] - min_bound[0])
+                    grid_ind1 = (np.floor(tmp_y)).astype(np.int)
+                    grid_ind2 = (np.floor((x_clip[:, 1:] - min_bound[1:]) / ((max_bound[1:] - min_bound[1:]) / (
+                                cur_grid_size[1:] - 1)))).astype(np.int)
+                    grid_ind = np.concatenate((grid_ind1.reshape((grid_ind1.shape[0], 1)), grid_ind2), axis=1)
+
+                    for i in range(5):
+                        left_x = i * 10
+                        right_x = (i + 1) * 10
+                        left_y = np.floor(a * quadratic_func(left_x))
+                        right_y = np.floor(a * quadratic_func(right_x))
+                        unique_cell = np.unique(grid_ind, axis=0)
+                        full_cell_num = np.sum(np.logical_and((unique_cell[:, 0] >= left_y), (unique_cell[:, 0] < right_y)))
+                        full_cell_list[i] += full_cell_num
+
+                elif func_type == "triple":
+                    x_clip = np.clip(xyz_pol, min_bound, max_bound)
+                    a = (cur_grid_size[0] - 1) / (triple_func(max_bound[0] - min_bound[0]))
+                    tmp_y = a * triple_func(x_clip[:, 0] - min_bound[0])
+                    grid_ind1 = (np.floor(tmp_y)).astype(np.int)
+                    grid_ind2 = (np.floor((x_clip[:, 1:] - min_bound[1:]) / ((max_bound[1:] - min_bound[1:]) / (
+                            cur_grid_size[1:] - 1)))).astype(np.int)
+                    grid_ind = np.concatenate((grid_ind1.reshape((grid_ind1.shape[0], 1)), grid_ind2), axis=1)
+
+                    for i in range(5):
+                        left_x = i * 10
+                        right_x = (i + 1) * 10
+                        left_y = np.floor(a * triple_func(left_x))
+                        right_y = np.floor(a * triple_func(right_x))
+                        unique_cell = np.unique(grid_ind, axis=0)
+                        full_cell_num = np.sum(np.logical_and((unique_cell[:, 0] >= left_y), (unique_cell[:, 0] < right_y)))
+                        full_cell_list[i] += full_cell_num
+
+                elif func_type == "ln":
+                    x_clip = np.clip(xyz_pol, min_bound, max_bound)
+                    a = (cur_grid_size[0] - 1) / (ln_func(max_bound[0] + 1 - min_bound[0]))
+                    tmp_y = a * ln_func(x_clip[:, 0] + 1 - min_bound[0])
+                    grid_ind1 = (np.floor(tmp_y)).astype(np.int)
+                    grid_ind2 = (np.floor((x_clip[:, 1:] - min_bound[1:]) / ((max_bound[1:] - min_bound[1:]) / (
+                            cur_grid_size[1:] - 1)))).astype(np.int)
+                    grid_ind = np.concatenate((grid_ind1.reshape((grid_ind1.shape[0], 1)), grid_ind2), axis=1)
+                    for i in range(5):
+                        left_x = i * 10
+                        right_x = (i + 1) * 10
+                        left_y = np.floor(a * ln_func(left_x))
+                        right_y = np.floor(a * ln_func(right_x))
+                        unique_cell = np.unique(grid_ind, axis=0)
+                        full_cell_num = np.sum(np.logical_and((unique_cell[:, 0] >= left_y), (unique_cell[:, 0] < right_y)))
+                        full_cell_list[i] += full_cell_num
+
+                elif func_type == "original":
+                    intervals = crop_range / (cur_grid_size - 1)
+                    if (intervals == 0).any(): print("Zero interval!")
+                    x_clip = np.clip(xyz_pol, min_bound, max_bound)
+                    grid_ind = (np.floor((x_clip - min_bound) / intervals)).astype(np.int)
+
+                    for i in range(5):
+                        left_x = i * 10
+                        right_x = (i + 1) * 10
+                        left_y = np.floor(left_x / intervals[0])
+                        right_y = np.floor(right_x / intervals[0])
+                        unique_cell = np.unique(grid_ind, axis=0)
+                        full_cell_num = np.sum(
+                            np.logical_and((unique_cell[:, 0] >= left_y), (unique_cell[:, 0] < right_y)))
+                        full_cell_list[i] += full_cell_num
+
+
+    for j in range(5):
+        left_x = j * 10
+        right_x = (j + 1) * 10
+        if func_type == "quadratic":
+            a = (cur_grid_size[0] - 1) / (quadratic_func(max_bound[0] - min_bound[0]))
+            left_y = np.floor(a * quadratic_func(left_x))
+            right_y = np.floor(a * quadratic_func(right_x))
+            total_cell_list[j] = (right_y - left_y) * dim2 * dim3 * total_file_num
+
+        elif func_type == "triple":
+            a = (cur_grid_size[0] - 1) / (triple_func(max_bound[0] - min_bound[0]))
+            left_y = np.floor(a * triple_func(left_x))
+            right_y = np.floor(a * triple_func(right_x))
+            total_cell_list[j] = (right_y - left_y) * dim2 * dim3 * total_file_num
+
+        elif func_type == "ln":
+            a = (cur_grid_size[0] - 1) / (ln_func(max_bound[0] + 1 - min_bound[0]))
+            left_y = np.floor(a * ln_func(left_x + 1))
+            right_y = np.floor(a * ln_func(right_x + 1))
+            total_cell_list[j] = (right_y - left_y) * dim2 * dim3 * total_file_num
+
+        elif func_type == "original":
+            intervals = crop_range / (cur_grid_size - 1)
+            left_y = np.floor((np.clip(left_x, min_bound[0], max_bound[0]) - min_bound[0]) / intervals[0])
+            right_y = np.floor((np.clip(right_x, min_bound[0], max_bound[0]) - min_bound[0]) / intervals[0])
+            total_cell_list[j] = (right_y - left_y) * dim2 * dim3 * total_file_num
+
+        res.append(full_cell_list[j] / total_cell_list[j])
+        print("res:", res)
+
+        with open(func_type + '.txt', 'w') as file:
+            for i in res:
+                file.write(str(i) + '\n')
+
+if __name__ == '__main__':
+    func_type = ["quadratic", "triple", "ln", "original"]
+    cal_cell_num(func_type[2])
diff --git a/train.sh b/train.sh
@@ -1,5 +1,5 @@
 name=cylinder_asym_networks
 gpuid=0
 
-CUDA_VISIBLE_DEVICES=${gpuid}  python -u train_cylinder_asym.py \
-2>&1 | tee logs_dir/${name}_logs_tee.txt
+CUDA_VISIBLE_DEVICES=${gpuid}  python3 -u train_cylinder_asym.py \
+2>&1 | tee logs_dir/${name}_logs_tee.txt