fix relative path on hubconfig to get cfg_file.

hubconf.py

@@ -1,6 +1,7 @@
 dependencies = ['torch', 'torchvision']
 
 import torch
+import os
 try:
   from mmcv.utils import Config, DictAction
 except:
@@ -41,14 +42,15 @@ def metric3d_convnext_large(pretrain=False, **kwargs):
   Returns:
     model (nn.Module): a Metric3D model.
   '''
-  cfg_file = MODEL_TYPE['ConvNeXt-Large']['cfg_file']
+  dirname = os.path.dirname(__file__)
+  cfg_file = os.path.join(dirname, MODEL_TYPE['ConvNeXt-Large']['cfg_file'])
   ckpt_file = MODEL_TYPE['ConvNeXt-Large']['ckpt_file']
 
   cfg = Config.fromfile(cfg_file)
   model = get_configured_monodepth_model(cfg)
   if pretrain:
     model.load_state_dict(
-      torch.hub.load_state_dict_from_url(ckpt_file)['model_state_dict'], 
+      torch.hub.load_state_dict_from_url(ckpt_file)['model_state_dict'],
       strict=False,
     )
   return model
@@ -62,14 +64,15 @@ def metric3d_vit_small(pretrain=False, **kwargs):
   Returns:
     model (nn.Module): a Metric3D model.
   '''
-  cfg_file = MODEL_TYPE['ViT-Small']['cfg_file']
+  dirname = os.path.dirname(__file__)
+  cfg_file = os.path.join(dirname, MODEL_TYPE['ViT-Small']['cfg_file'])
   ckpt_file = MODEL_TYPE['ViT-Small']['ckpt_file']
 
   cfg = Config.fromfile(cfg_file)
   model = get_configured_monodepth_model(cfg)
   if pretrain:
     model.load_state_dict(
-      torch.hub.load_state_dict_from_url(ckpt_file)['model_state_dict'], 
+      torch.hub.load_state_dict_from_url(ckpt_file)['model_state_dict'],
       strict=False,
     )
   return model
@@ -83,14 +86,15 @@ def metric3d_vit_large(pretrain=False, **kwargs):
   Returns:
     model (nn.Module): a Metric3D model.
   '''
-  cfg_file = MODEL_TYPE['ViT-Large']['cfg_file']
+  dirname = os.path.dirname(__file__)
+  cfg_file = os.path.join(dirname, MODEL_TYPE['ViT-Large']['cfg_file'])
   ckpt_file = MODEL_TYPE['ViT-Large']['ckpt_file']
 
   cfg = Config.fromfile(cfg_file)
   model = get_configured_monodepth_model(cfg)
   if pretrain:
     model.load_state_dict(
-      torch.hub.load_state_dict_from_url(ckpt_file)['model_state_dict'], 
+      torch.hub.load_state_dict_from_url(ckpt_file)['model_state_dict'],
       strict=False,
     )
   return model
@@ -104,14 +108,15 @@ def metric3d_vit_giant2(pretrain=False, **kwargs):
   Returns:
     model (nn.Module): a Metric3D model.
   '''
-  cfg_file = MODEL_TYPE['ViT-giant2']['cfg_file']
+  dirname = os.path.dirname(__file__)
+  cfg_file = os.path.join(dirname, MODEL_TYPE['ViT-giant2']['cfg_file'])
   ckpt_file = MODEL_TYPE['ViT-giant2']['ckpt_file']
 
   cfg = Config.fromfile(cfg_file)
   model = get_configured_monodepth_model(cfg)
   if pretrain:
     model.load_state_dict(
-      torch.hub.load_state_dict_from_url(ckpt_file)['model_state_dict'], 
+      torch.hub.load_state_dict_from_url(ckpt_file)['model_state_dict'],
       strict=False,
     )
   return model
@@ -163,7 +168,7 @@ def metric3d_vit_giant2(pretrain=False, **kwargs):
   # un pad
   pred_depth = pred_depth.squeeze()
   pred_depth = pred_depth[pad_info[0] : pred_depth.shape[0] - pad_info[1], pad_info[2] : pred_depth.shape[1] - pad_info[3]]
-  
+
   # upsample to original size
   pred_depth = torch.nn.functional.interpolate(pred_depth[None, None, :, :], rgb_origin.shape[:2], mode='bilinear').squeeze()
   ###################### canonical camera space ######################
@@ -173,14 +178,14 @@ def metric3d_vit_giant2(pretrain=False, **kwargs):
   pred_depth = pred_depth * canonical_to_real_scale # now the depth is metric
   pred_depth = torch.clamp(pred_depth, 0, 300)
 
-  #### you can now do anything with the metric depth 
+  #### you can now do anything with the metric depth
   # such as evaluate predicted depth
   if depth_file is not None:
     gt_depth = cv2.imread(depth_file, -1)
     gt_depth = gt_depth / gt_depth_scale
     gt_depth = torch.from_numpy(gt_depth).float().cuda()
     assert gt_depth.shape == pred_depth.shape
-    
+
     mask = (gt_depth > 1e-8)
     abs_rel_err = (torch.abs(pred_depth[mask] - gt_depth[mask]) / gt_depth[mask]).mean()
     print('abs_rel_err:', abs_rel_err.item())

mono/configs/HourglassDecoder/convlarge.0.3_150.py

@@ -17,7 +17,7 @@
         focal_length=1000.0,
     ),
     depth_range=(0, 1),
-    depth_normalize=(0.3, 150),
+    depth_normalize=(0.3, 500),
     crop_size = (544, 1216),
 ) 
 

mono/configs/HourglassDecoder/vit.raft5.giant2.py

@@ -15,7 +15,7 @@
 )
 
 
-max_value = 200
+max_value = 500
 # configs of the canonical space
 data_basic=dict(
     canonical_space = dict(
@@ -25,7 +25,7 @@
     depth_range=(0, 1),
     depth_normalize=(0.1, max_value),
     crop_size = (616, 1064),  # %28 = 0
-     clip_depth_range=(0.1, 200),
+     clip_depth_range=(0.1, 500),
     vit_size=(616,1064)
 ) 
 

mono/model/decode_heads/RAFTDepthNormalDPTDecoder5.py

@@ -216,7 +216,7 @@ def compute_depth_expectation(prob, depth_values):
     return depth
 
 def interpolate_float32(x, size=None, scale_factor=None, mode='nearest', align_corners=None):
-    with torch.autocast(device_type='cuda', dtype=torch.bfloat16, enabled=False):
+    with torch.autocast(device_type='cuda', dtype=torch.float, enabled=False):
         return F.interpolate(x.float(), size=size, scale_factor=scale_factor, mode=mode, align_corners=align_corners)
 
 # def upflow8(flow, mode='bilinear'):
@@ -225,7 +225,7 @@ def interpolate_float32(x, size=None, scale_factor=None, mode='nearest', align_c
 
 def upflow4(flow, mode='bilinear'):
     new_size = (4 * flow.shape[2], 4 * flow.shape[3])
-    with torch.autocast(device_type='cuda', dtype=torch.bfloat16, enabled=False):
+    with torch.autocast(device_type='cuda', dtype=torch.float, enabled=False):
         return  F.interpolate(flow, size=new_size, mode=mode, align_corners=True)
 
 def coords_grid(batch, ht, wd):