allow fix number of estimated points

javiribera · Feb 3, 2018 · 2539326 · 2539326
1 parent 512e8a7
commit 2539326
Show file tree

Hide file tree

Showing 3 changed files with 33 additions and 22 deletions.
diff --git a/test.py b/test.py
@@ -123,16 +123,6 @@ def __getitem__(self, idx):
                                  batch_size=args.eval_batch_size,
                                  num_workers=args.nThreads)
 
-# Model
-print('Building network... ', end='')
-# model = unet.UnetGenerator(input_nc=3, output_nc=1, num_downs=8)
-model = unet_model.UNet(3, 1, known_n_points=args.n_points)
-print('DONE')
-print(model)
-model = nn.DataParallel(model)
-if args.cuda:
-    model.cuda()
-
 # Loss function
 l1_loss = nn.L1Loss()
 criterion_training = losses.ModifiedChamferLoss(256, 256, return_2_terms=True)
@@ -142,9 +132,28 @@ def __getitem__(self, idx):
 if os.path.isfile(args.checkpoint):
     checkpoint = torch.load(args.checkpoint)
     start_epoch = checkpoint['epoch']
+    # Model
+    if args.n_points is None:
+        if 'n_points' not in checkpoint:
+            # Model will also estimate # of points
+            model=unet_model.UNet(3, 1, None)
+        else:
+            # The checkpoint tells us the # of points to estimate
+            model=unet_model.UNet(3, 1, checkpoint['n_points'])
+    else:
+        # The user tells us the # of points to estimate
+        model=unet_model.UNet(3, 1, known_n_points=args.n_points)
+
+    # Parallelize
+    model = nn.DataParallel(model)
+    if args.cuda:
+        model.cuda()
+
+    # Load model in checkpoint
     model.load_state_dict(checkpoint['model'])
     print("╰─ loaded checkpoint '{}' (now on epoch {})"
           .format(args.checkpoint, checkpoint['epoch']))
+    print(model)
 else:
     print("╰─ E: no checkpoint found at '{}'".format(args.checkpoint))
     exit(-1)

diff --git a/train_and_validate.py b/train_and_validate.py
@@ -454,6 +454,7 @@ def __getitem__(self, idx):
                         'model': model.state_dict(),
                         'lowest_avg_ahd_val': avg_ahd_val_float,
                         'optimizer': optimizer.state_dict(),
+                        'n_points': args.n_points,
                         }, args.save)
             print("Saved best checkpoint so far in %s " % args.save)
 

diff --git a/unet_model.py b/unet_model.py
@@ -7,7 +7,8 @@
 
 
 class UNet(nn.Module):
-    def __init__(self, n_channels, n_classes, known_n_points=None):
+    def __init__(self, n_channels, n_classes,
+                 known_n_points=None):
         super(UNet, self).__init__()
         self.inc = inconv(n_channels, 64)
         self.down1 = down(64, 128)
@@ -29,12 +30,10 @@ def __init__(self, n_channels, n_classes, known_n_points=None):
         self.outc = outconv(64, n_classes)
         self.out_nonlin = nn.Sigmoid()
 
-        self.regressor = nn.Linear(256*256, 1)
-        self.regressor_nonlin = nn.Softplus()
-
-        self.lin = nn.Linear(1, 1, bias=False)
-
         self.known_n_points = known_n_points
+        if known_n_points is None:
+            self.regressor = nn.Linear(288*384, 1)
+            self.regressor_nonlin = nn.Softplus()
 
     def forward(self, x):
         x1 = self.inc(x)
@@ -57,12 +56,14 @@ def forward(self, x):
         x = self.outc(x)
         x = self.out_nonlin(x)
 
-        x_flat = x.view(1, -1)
-
-        regression = self.regressor(x_flat)
-        regression = self.regressor_nonlin(regression)
-
-        return x, regression
+        if self.known_n_points is None:
+            x_flat = x.view(1, -1)
+            regression = self.regressor(x_flat)
+            regression = self.regressor_nonlin(regression)
+            return x, regression
+        else:
+            n_pts = Variable(torch.cuda.FloatTensor([self.known_n_points]))
+            return x, n_pts
         # summ = torch.sum(x)
         # count = self.lin(summ)