支持是否使用pretrain

img_dataset.py

@@ -163,7 +163,62 @@ def crop_img():
         region.save(croped_img_file)
 
 
+def stats():
+    excel_name_cnt = 0
+    excel_2_file_name_cnt = 0
+    rootdir = 'E:\\workspace\\zhongyi'
+    dirs = os.listdir(rootdir)  # 列出文件夹下所有的目录与文件
+    for i in range(0, len(dirs)):
+        path = os.path.join(rootdir, dirs[i])
+        if os.path.isdir(path):
+            # take image dir and xlsx file
+            sub_paths = os.listdir(path)
+            img_paths = []
+            xlsx_path = None
+            for j in range(0, len(sub_paths)):
+                sub_path = os.path.join(path, sub_paths[j])
+                if os.path.isdir(sub_path):
+                    img_paths.append(sub_path)
+                elif os.path.isfile(sub_path):
+                    xlsx_path = sub_path
+        img_files=[]
+        for img_path in img_paths:
+            cur_dir_img_files = os.listdir(img_path)
+            img_files.extend(cur_dir_img_files)
+        print('在[{0}]下总共有[{1}]张图片'.format(path, len(img_files)))
+        # 根据excel的病人名字去sub_path找对应的名字
+        patient_info = pd.read_excel(xlsx_path, sheet_name=None)
+        patient_names=[]
 
-if __name__ == "__main__":
-    preprocess_data()
-    # crop_img()
+        for sheet_name, sheet in patient_info.items():
+            is_find_name = False
+            for col_nm1 in sheet.columns:
+                if '姓名' in col_nm1:
+                    is_find_name=True
+                    cur_clo_names=sheet[col_nm1].values
+                    # print('-->在[{0}].[{1}].[{2}]找到[{3}]人'.format(xlsx_path, sheet_name, col_nm1, len(cur_clo_names)))
+                    patient_names.extend(cur_clo_names)
+            if not is_find_name:
+                print('WARNING, 在[{0}].[{1}].[{2}]没找到包含[姓名]字眼的属性'.format(xlsx_path, sheet_name, col_nm1))
+        patient_names=list(set(patient_names))
+        print('-->在[{0}]找到[{1}]人'.format(xlsx_path, len(patient_names)))
+        find_name_cnt = 0
+        for name in patient_names:
+            if isinstance(name, float):
+                # print('error name:', name)
+                continue
+            for img in img_files:
+                if str(name) in img:
+                    find_name_cnt += 1
+                    break
+        print('-->根据excel姓名找图找到{0}/{1}人'.format(find_name_cnt, len(patient_names)))
+        excel_name_cnt += len(patient_names)
+        excel_2_file_name_cnt += find_name_cnt
+        print()
+    print('-->汇总：根据excel姓名找图找到{0}/{1}人'.format(excel_2_file_name_cnt, excel_name_cnt))
+
+
+# if __name__ == "__main__":
+#     # preprocess_data()
+#     # crop_img()
+#     stats()

train_tongue_diagnosis.py

@@ -8,7 +8,8 @@ import numpy as np
 import torch.nn.functional as F
 
 from img_dataset import Image_DataSet
-from models import wideresnet, resnet
+# from torch.autograd import Variable
+# from graphviz import Digraph
 
 from torchvision import models
 
@@ -22,9 +23,9 @@ parser.add_argument('--lr', type=float, default=0.001, metavar='LR', help='learn
 parser.add_argument('--momentum', type=float, default=0.9, metavar='M', help='SGD momentum')
 parser.add_argument('--weight_decay', '--wd', default=0, type=float, metavar='W')
 parser.add_argument('--epochs', type=int, default=40, metavar='N', help='number of epochs to train')
-parser.add_argument('--schedule', type=int, nargs='+', default=[25, 30, 35],
+parser.add_argument('--schedule', type=int, nargs='+', default=[10, 20, 30],
                     help='Decrease learning rate at these epochs.')
-parser.add_argument('--gamma', type=float, default=0.2, help='LR is multiplied by gamma on schedule.')
+parser.add_argument('--gamma', type=float, default=0.5, help='LR is multiplied by gamma on schedule.')
 parser.add_argument('--model_dir', default='dnn_models/tongue_modes/shetai/', help='directory of model for saving checkpoint')
 parser.add_argument('--training_data_dir', default='./processed-data-shetai', help='directory of training data')
 parser.add_argument('--test_data_dir', default='./shetai-test', help='directory of test data')
@@ -32,6 +33,8 @@ parser.add_argument('--save_freq', '-s', default=1, type=int, metavar='N', help=
 parser.add_argument('--log_interval', type=int, default=10, metavar='N',
                     help='how many batches to wait before logging training status')
 parser.add_argument('--seed', type=int, default=1, metavar='S', help='random seed (default: 1)')
+parser.add_argument('--pretrain', action='store_true', default=True,
+                    help='whether to use pretrain')
 
 args = parser.parse_args()
 
@@ -83,15 +86,15 @@ tongue_test_loader = torch.utils.data.DataLoader(tongue_dataset_test, batch_size
 
 def get_model(model_name):
     if model_name == 'densenet121':
-        model = models.densenet121(pretrained=True)
+        model = models.densenet121(pretrained=args.pretrain)
         model.classifier = torch.nn.Linear(in_features=1024, out_features=NUM_CLASSES, bias=True)
         return model.to(device)
     elif model_name == 'resnet-50':
-        model = models.resnet50(pretrained=True)
+        model = models.resnet50(pretrained=args.pretrain)
         model.fc = torch.nn.Linear(in_features=2048, out_features=NUM_CLASSES, bias=True)
         return model.to(device)
     elif model_name == 'vgg19_bn':
-        model = models.vgg19_bn(pretrained=True)
+        model = models.vgg19_bn(pretrained=args.pretrain)
         model.classifier[6] = torch.nn.Linear(in_features=4096, out_features=NUM_CLASSES, bias=True)
         return model.to(device)
     else:
@@ -171,6 +174,77 @@ def eval_test(model, test_loader):
     return test_loss, test_accuracy
 
 
+def make_dot(var, params=None):
+    """ Produces Graphviz representation of PyTorch autograd graph
+    Blue nodes are the Variables that require grad, orange are Tensors
+    saved for backward in torch.autograd.Function
+    Args:
+        var: output Variable
+        params: dict of (name, Variable) to add names to node that
+            require grad (TODO: make optional)
+    """
+    if params is not None:
+        assert isinstance(params.values()[0], Variable)
+        param_map = {id(v): k for k, v in params.items()}
+
+    node_attr = dict(style='filled',
+                     shape='box',
+                     align='left',
+                     fontsize='12',
+                     ranksep='0.1',
+                     height='0.2')
+    dot = Digraph(node_attr=node_attr, graph_attr=dict(size="12,12"))
+    seen = set()
+
+    def size_to_str(size):
+        return '('+(', ').join(['%d' % v for v in size])+')'
+
+    def add_nodes(var):
+        if var not in seen:
+            if torch.is_tensor(var):
+                dot.node(str(id(var)), size_to_str(var.size()), fillcolor='orange')
+            elif hasattr(var, 'variable'):
+                u = var.variable
+                name = param_map[id(u)] if params is not None else ''
+                node_name = '%s\n %s' % (name, size_to_str(u.size()))
+                dot.node(str(id(var)), node_name, fillcolor='lightblue')
+            else:
+                dot.node(str(id(var)), str(type(var).__name__))
+            seen.add(var)
+            if hasattr(var, 'next_functions'):
+                for u in var.next_functions:
+                    if u[0] is not None:
+                        dot.edge(str(id(u[0])), str(id(var)))
+                        add_nodes(u[0])
+            if hasattr(var, 'saved_tensors'):
+                for t in var.saved_tensors:
+                    dot.edge(str(id(t)), str(id(var)))
+                    add_nodes(t)
+    add_nodes(var.grad_fn)
+    return dot
+
+
+def print_model_arch(model):
+    x = torch.randn(1, 3, 298, 224).to(device)
+    y = model(x)
+    g = make_dot(y)
+    g.view()
+
+    # from torchsummary import summary
+    # summary(model, (3, 224, 224))
+
+    params = list(model.parameters())
+    k = 0
+    for i in params:
+        l = 1
+        print("该层的结构：" + str(list(i.size())))
+        for j in i.size():
+            l *= j
+        print("该层参数和：" + str(l))
+        k = k + l
+    print("总参数数量和：" + str(k))
+
+
 def main():
     print(args)
     model = get_model(args.model_name).to(device)