display loss in train function

loss.py

@@ -1,8 +1,3 @@
-'''
-Reference :
-   https://github.com/kuangliu/pytorch-retinanet/blob/master/loss.py
-'''
-
 from __future__ import print_function
 
 import torch
@@ -33,7 +28,6 @@ def focal_loss2d(self, x, y):
             t = one_hot_embedding(y.data.cpu(), 2)
         else:
             t = one_hot_embedding(y.data, 2)
-        # t = t[:,1:]  # exclude background
 
         if self.using_gpu is True:
             t = Variable(t).cuda()
@@ -135,11 +129,9 @@ def forward(self, loc_preds, loc_targets, cls_preds, cls_targets, mask_preds, ma
         masked_cls_preds = cls_preds[mask].view(-1,self.num_classes)
         cls_loss = self.focal_loss(masked_cls_preds, cls_targets[pos_neg])
 
-        # mask_loss = self.ce_loss(mask_preds, mask_targets)
-        mask_loss = self.focal_loss2d(mask_preds, mask_targets)
+        mask_loss = self.ce_loss(mask_preds, mask_targets)
+        # mask_loss = self.focal_loss2d(mask_preds, mask_targets)
 
 
-        print('loc_loss: %.3f | cls_loss: %.3f | mask_loss: %.3f' %
-              (loc_loss.data[0]/num_pos, cls_loss.data[0]/num_pos, mask_loss.data[0]), end=' | ')
-        loss = ((loc_loss+cls_loss)/num_pos) + (mask_loss)
-        return loss, num_pos
+        # loss = ((loc_loss+cls_loss)/num_pos) + (mask_loss)
+        return loc_loss, cls_loss, mask_loss, num_pos

train.py

@@ -50,7 +50,7 @@
 
 validset = ListDataset(root="../valid", gt_extension=".txt",
                       labelmap_path="class_label_map.xlsx", is_train=False, transform=transform, input_image_size=512,
-                      num_crops=1, original_img_size=512)
+                      num_crops=5, original_img_size=512)
 validloader = torch.utils.data.DataLoader(validset, batch_size=batch_size, shuffle=False, num_workers=8, collate_fn=validset.collate_fn)
 
 print("lr : " + str(lr))
@@ -130,14 +130,17 @@ def train(epoch):
 
         optimizer.zero_grad()
         loc_preds, cls_preds, mask_preds = net(inputs)
-        loss, num_matched_anchors = criterion(loc_preds, loc_targets, cls_preds, cls_targets, mask_preds, mask_targets)
+        loc_loss, cls_loss, mask_loss, num_matched_anchors = \
+            criterion(loc_preds, loc_targets, cls_preds, cls_targets, mask_preds, mask_targets)
+        loss = ((loc_loss + cls_loss) / num_matched_anchors) + mask_loss
         loss.backward()
         optimizer.step()
 
         train_loss += loss.data[0]
         avg_matched_anchor += float(num_matched_anchors)
-        print('epoch: %3d | iter: %4d | train_loss: %.3f | avg_loss: %.3f | avg_num. matched: %d'
-              % (epoch, batch_idx, loss.data[0], train_loss/(batch_idx+1), avg_matched_anchor/(batch_idx+1)))
+        print('epoch: %3d | iter: %4d | loc_loss: %.3f | cls_loss: %.3f | mask_loss: %.3f | train_loss: %.3f | avg_loss: %.3f | avg_num. matched: %d'
+              % (epoch, batch_idx, loc_loss.data[0] / num_matched_anchors, cls_loss.data[0] / num_matched_anchors,
+                 mask_loss.data[0], loss.data[0], train_loss / (batch_idx + 1), avg_matched_anchor / (batch_idx + 1)))
 
 # Test
 def valid(epoch):
@@ -158,11 +161,17 @@ def valid(epoch):
             mask_targets = Variable(mask_targets)
 
         loc_preds, cls_preds, mask_preds = net(inputs)
-        loss, _ = criterion(loc_preds, loc_targets, cls_preds, cls_targets, mask_preds, mask_targets)
+        loc_loss, cls_loss, mask_loss, num_matched_anchors = \
+            criterion(loc_preds, loc_targets, cls_preds, cls_targets, mask_preds, mask_targets)
+        loss = ((loc_loss + cls_loss) / num_matched_anchors) + mask_loss
         valid_loss += loss.data[0]
-        print('valid_loss: %.3f | avg_loss: %.3f' % (loss.data[0], valid_loss/(batch_idx+1)))
+        print('loc_loss: %.3f | cls_loss: %.3f | valid_loss: %.3f | avg_loss: %.3f'
+              % (loc_loss.data[0] / num_matched_anchors, cls_loss.data[0] / num_matched_anchors,
+                 loss.data[0], valid_loss / (batch_idx + 1)))
 
     # Save checkpoint
+    # Every checkpoints are stored to analyze how is going training
+    # Model is selected by low-validation error.
     valid_loss /= len(validloader)
     print('Saving..')
     state = {