Refactoring transfer learning code.

pysegcnn/main/train_transfer.py

@@ -101,7 +101,7 @@ if __name__ == '__main__':
             # transfer learning
             net, optimizer, checkpoint = trn_sf.transfer_model(
                 trn_sf.pretrained_path,
-                nclasses=len(src_ds).labels,
+                nclasses=len(src_ds.labels),
                 optim_kwargs=net_mc.optim_kwargs,
                 freeze=trn_sf.freeze)
         else:

pysegcnn/main/train_transfer_config.py

@@ -29,10 +29,14 @@ HERE = pathlib.Path(__file__).resolve().parent
 # path to the datasets on the current machine
 DRIVE_PATH = pathlib.Path('C:/Eurac/Projects/CCISNOW/Datasets/')
 # DRIVE_PATH = pathlib.Path('/mnt/CEPH_PROJECTS/cci_snow/dfrisinghelli/Datasets/')  # nopep8
+# DRIVE_PATH = pathlib.Path('/home/dfrisinghelli/Datasets/')
+# DRIVE_PATH = pathlib.Path('/home/clusterusers/dfrisinghelli_eurac/Datasets/')
+# DRIVE_PATH = pathlib.Path('/scratch/dfrisinghelli_eurac/Datasets/')
+# DRIVE_PATH = pathlib.Path('/localscratch/dfrisinghelli_eurac/Datasets/')
 
 # name and paths to the datasets
 DATASETS = {'Sparcs': DRIVE_PATH.joinpath('Sparcs'),
-            'Alcd': DRIVE_PATH.joinpath('Alcd/60m')
+            'Alcd': DRIVE_PATH.joinpath('Alcd')
             }
 
 # name of the source domain dataset
@@ -45,10 +49,10 @@ TRG_DS = 'Alcd'
 BANDS = ['red', 'green', 'blue', 'nir', 'swir1', 'swir2']
 
 # tile size of a single sample
-TILE_SIZE = 128
+TILE_SIZE = 64
 
 # number of folds for cross validation
-K_FOLDS = 2
+K_FOLDS = 1
 
 # the source dataset configuration dictionary
 src_ds_config = {
@@ -206,7 +210,7 @@ src_split_config = {
     # (ttratio * tvratio) * 100 % will be used for training
     # (1 - ttratio * tvratio) * 100 % will be used for validation
     # used if 'kfolds=1'
-    'tvratio': 0.8,
+    'tvratio': 0.05,
 
 }
 
@@ -219,7 +223,7 @@ trg_split_config = {
     'seed': 0,
     'shuffle': True,
     'ttratio': 1,
-    'tvratio': 0.8,
+    'tvratio': 0.05,
 
 }
 
@@ -281,7 +285,7 @@ model_config = {
 
     # define the number of epochs: the number of maximum iterations over
     # the whole training dataset
-    'epochs': 100,
+    'epochs': 10,
 
 }
 
@@ -294,8 +298,8 @@ tlda_config = {
 
     # whether to apply any sort of transfer learning
     # if transfer=False, the model is only trained on the source dataset
-    # 'transfer': True,
-    'transfer': False,
+    'transfer': True,
+    # 'transfer': False,
 
     # Supervised vs. Unsupervised ---------------------------------------------
     # -------------------------------------------------------------------------
@@ -313,13 +317,13 @@ tlda_config = {
     #                   scratch ('uda_from_pretrained=False') or the pretrained
     #                   model in 'pretrained_model' is loaded
     #                   ('uda_from_pretrained=True')
-    # 'supervised': True,
-    'supervised': False,
+    'supervised': True,
+    # 'supervised': False,
 
     # name of the pretrained model to apply for transfer learning
     # required if transfer=True and supervised=True
     # optional if transfer=True and supervised=False
-    'pretrained_model': '',  # nopep8
+    'pretrained_model': 'Segnet_Adam_b128_AlcdDataset_m2_Scene_s0t10v08_t64_b2g3r4.pt',  # nopep8
 
     # loss function for unsupervised domain adaptation
     # currently supported methods:
@@ -356,6 +360,7 @@ tlda_config = {
     # 'uda_pos': 'cla',
 
     # whether to freeze the pretrained model weights
-    'freeze': True,
+    # 'freeze': True
+    'freeze': False
 
 }