1

app.py

@@ -92,7 +92,16 @@ def generate_monocular_depth_maps(img_list, depth_prior_name):
     focallength_px_list = []
     
     if depth_prior_name=='Depth Pro':
-        model, transform = depth_pro.create_model_and_transforms(device='cuda')
+        
+        DEFAULT_MONODEPTH_CONFIG_DICT = DepthProConfig(
+            patch_encoder_preset="dinov2l16_384",
+            image_encoder_preset="dinov2l16_384",
+            checkpoint_uri=f'{HERE_PATH}/third_party/ml-depth-pro/checkpoints/depth_pro.pt',
+            decoder_features=256,
+            use_fov_head=True,
+            fov_encoder_preset='dinov2l16_384',
+        )
+        model, transform = depth_pro.create_model_and_transforms(config=DEFAULT_MONODEPTH_CONFIG_DICT,  device='cuda')
         model.eval()
 
         for image_path in tqdm(img_list):

dust3r/cloud_opt_flow/base_opt.py

@@ -28,7 +28,7 @@ from PIL import Image
 import os.path as path
 HERE_PATH = path.normpath(path.dirname(__file__)) 
 print('**',HERE_PATH)
-
+#** /home/user/app/dust3r/cloud_opt_flow
 def c2w_to_tumpose(c2w):
     """
     Convert a camera-to-world matrix to a tuple of translation and rotation

third_party/ml-depth-pro/src/depth_pro/depth_pro.py

@@ -40,7 +40,7 @@ class DepthProConfig:
 DEFAULT_MONODEPTH_CONFIG_DICT = DepthProConfig(
     patch_encoder_preset="dinov2l16_384",
     image_encoder_preset="dinov2l16_384",
-    checkpoint_uri="/home/lipeng/ljh_code/Video_Depth_CVPR2025-main/ml-depth-pro/checkpoints/depth_pro.pt",
+    checkpoint_uri="./third_party/ml-depth-pro/checkpoints/depth_pro.pt",
     decoder_features=256,
     use_fov_head=True,
     fov_encoder_preset="dinov2l16_384",