Training

train.py

@@ -24,22 +24,22 @@
 transforms_compose = transforms.Compose(
     [transforms.ToTensor()])
 params = hyperparameters(
-    train_percentage=0.6,
+    train_percentage=1,
     batch_size=1,
-    epoch=200,
-    n_classes=1)
+    epoch=50,
+    n_classes=2)
 
 if torch.cuda.is_available():
-    net = UNeT(n_classes=29, n_channels=3).cuda()
+    net = UNeT(n_classes=2, n_channels=3).cuda()
 else:
-    net = UNeT(n_classes=29, n_channels=3)
+    net = UNeT(n_classes=2, n_channels=3)
 encoder= HotEncoder(is_binary= False, dir= ANNOTATIONS_DIR, extension="png")
+color_dict= encoder.gen_colors()
 Images = ImageList(
     Images=IMAGE_DIR,
     Annotations=ANNOTATIONS_DIR,
     train_percentage=1,
     extension="png")
-loss_val = Loss()
 Train = ImageLoader(
     encoder_obj= encoder,
     data=Images.train_set,
@@ -58,7 +58,6 @@
     Test,
     batch_size=params.hyperparameters["batch_size"],
     shuffle=True)
-color_dict= encoder.gen_colors()
 criterion = nn.BCEWithLogitsLoss()
 optimizer = optim.SGD(net.parameters(), lr=0.001, momentum=0.9)
 for i, data in enumerate(TrainLoader, 0):
@@ -68,6 +67,7 @@
     metrics = defaultdict()
     running_loss = 0.0
     for i, data in enumerate(TrainLoader, 0):
+        net.train()
         if torch.cuda.is_available():
             inputs, labels = data["Image"].cuda(), data["Label"].cuda()
         else:
@@ -80,11 +80,7 @@
         loss.backward()
         optimizer.step()
         running_loss += loss.item()
-        print(
-            "Epoch: {} | Loss: {} | Instance: {}".format(
-                int(epoch),
-                loss.item(),
-                i))
-        print("Running loss|", running_loss)
-torch.save(net.state_dict(), MODEL_SAVE + "/model.pt")
-print("Model saved")
+        print(loss)
+torch.save(net.state_dict(), MODEL_SAVE + "/model_roof_epoch50_batch_5.pt")
+
+print("Model saved")