add : new dataset

config/config.py

@@ -20,6 +20,7 @@ def load_args():
     parser.add_argument('--save_path', type=str, default='output/best_model.pt')
     parser.add_argument('--pretrain_path', type=str, default=None)
     parser.add_argument('--pretrain_imgnet', action=argparse.BooleanOptionalAction, default=False)
+    parser.add_argument('--resolution', type=str, default=None)
     args = parser.parse_args()
 
     return args

main.py

@@ -194,18 +194,23 @@ def test_duo(model, data_test, loss_function, epoch):
 def run_duo(args):
     #load data
 
-    #data_train, data_test,_ = load_data_duo(base_dir_train=args.dataset_train_dir, base_dir_val=args.dataset_val_dir, base_dir_test=args.dataset_test_dir, batch_size=args.batch_size)
-    data_train, data_test = load_data_duo_patch(base_dir_train=args.dataset_train_dir, base_dir_val=args.dataset_val_dir,
-                                             base_dir_test=args.dataset_test_dir, batch_size=args.batch_size)
+    if args.resolution == 'max':
+        data_train, data_test = load_data_duo_patch(base_dir_train=args.dataset_train_dir,
+                                                    base_dir_val=args.dataset_val_dir,
+                                                    base_dir_test=args.dataset_test_dir, batch_size=args.batch_size)
+    else :
+        data_train, data_test,_ = load_data_duo(base_dir_train=args.dataset_train_dir, base_dir_val=args.dataset_val_dir, base_dir_test=args.dataset_test_dir, batch_size=args.batch_size)
+
     #load model
     if not args.pretrain_imgnet:
         #adaptative path model
         for imaer,imana, label in data_train:
             img_shape = imaer.shape # batch x tile x img
             break
-        model = Classification_model_duo_tile(model = args.model, n_class=len(data_train.dataset.classes),n_tile=img_shape[1])
-
-        #model = Classification_model_duo(model = args.model, n_class=len(data_train.dataset.classes))
+        if args.resolution == 'max' :
+            model = Classification_model_duo_tile(model = args.model, n_class=len(data_train.dataset.classes),n_tile=img_shape[1])
+        else :
+            model = Classification_model_duo(model = args.model, n_class=len(data_train.dataset.classes))
     else :
         model = Classification_model_duo_pretrained(model = args.model, n_class=len(data_train.dataset.classes))