dimension fix

.gitignore

@@ -5,4 +5,4 @@ logs/
 results/
 __pycache__/
 code/dataset/filenames/nyudepthv2/
-code/dataset/weights/
+code/dataset/weights/*

code/dataset/base_dataset.py

@@ -37,6 +37,7 @@ def readTXT(self, txt_path):
 
     def augment_training_data(self, image, depth):
         H, W, C = image.shape
+
         if self.count % 4 == 0:
             alpha = random.random()
             beta = random.random()
@@ -68,4 +69,3 @@ def augment_test_data(self, image, depth):
         depth = self.to_tensor(depth).squeeze()
 
         return image, depth
-

code/dataset/nyudepthv2.py

@@ -48,8 +48,8 @@ def __getitem__(self, idx):
 
         H, W, C = image.shape
         #first, image and depth must be cropped because of the large borders on the depth map
-        image=image[int(W/10),int(W*19/20),int(W/12):int(W*11/12)]
-        depth=depth[int(W/10),int(W*19/20),int(W/12):int(W*11/12)]
+        image=image[40:460,40:600,:]
+        depth=depth[40:460,40:600]
 
         if self.is_train and self.do_cutdepth:
             image, depth = self.augment_training_data(image, depth)

code/dataset/sunrgbd.py

@@ -44,11 +44,6 @@ def __getitem__(self, idx):
         if self.scale_size:
             image = cv2.resize(image, (self.scale_size[1], self.scale_size[0]))
             depth = cv2.resize(depth, (self.scale_size[1], self.scale_size[0]))
-        H, W, C = image.shape
-        #first, image and depth must be cropped because of the large borders on the depth map
-        image=image[int(W/10),int(W*19/20),int(W/12):int(W*11/12)]
-        depth=depth[int(W/10),int(W*19/20),int(W/12):int(W*11/12)]
-
         if self.is_train and self.do_cutdepth:
             image, depth = self.augment_training_data(image, depth)
         else:

code/models/model.py

@@ -10,17 +10,17 @@ def __init__(self, max_depth=10.0, is_train=False):
         self.max_depth = max_depth
 
         self.encoder = mit_b4()
-        if is_train:            
-            ckpt_path = './GLPDepth/code/models/weights/mit_b4.pth'
-            try:
-                load_checkpoint(self.encoder, ckpt_path, logger=None)
-            except:
-                import gdown
-                print("Download pre-trained encoder weights...")
-                id = '1BUtU42moYrOFbsMCE-LTTkUE-mrWnfG2'
-                url = 'https://drive.google.com/uc?id=' + id
-                output = './GLPDepth/code/models/weights/mit_b4.pth'
-                gdown.download(url, output, quiet=False)
+        # if is_train:            
+        #     ckpt_path = './GLPDepth/code/models/weights/mit_b4.pth'
+        #     try:
+        #         load_checkpoint(self.encoder, ckpt_path, logger=None)
+        #     except:
+        #         import gdown
+        #         print("Download pre-trained encoder weights...")
+        #         id = '1BUtU42moYrOFbsMCE-LTTkUE-mrWnfG2'
+        #         url = 'https://drive.google.com/uc?id=' + id
+        #         output = './GLPDepth/code/models/weights/mit_b4.pth'
+        #         gdown.download(url, output, quiet=False)
 
         channels_in = [512, 320, 128]
         channels_out = 64

code/train.py

@@ -57,9 +57,9 @@ def main():
     start=time.time()
 
     # Dataset setting
-    dataset_kwargs = {'dataset_name': args.dataset, 'data_path': args.data_path,'filenames_path':args.filenames_path}
+    dataset_kwargs = {'dataset_name': args.dataset, 'data_path': args.data_path,'filenames_path':args.filenames_path,'do_cutdepth':args.do_cutdepth}
     if args.dataset == 'nyudepthv2':
-        dataset_kwargs['crop_size'] = (448, 576)
+        dataset_kwargs['crop_size'] = (448-32,576-32)
     elif args.dataset == 'kitti':
         dataset_kwargs['crop_size'] = (352, 704)
     else: