MCPcopy Create free account
hub / github.com/MegaScenes/nvs / load_depth_model

Function load_depth_model

dataloader/evalhelpers.py:118–137  ·  view source on GitHub ↗
()

Source from the content-addressed store, hash-verified

116
117
118def load_depth_model():
119 # load depth model
120 with warnings.catch_warnings():
121 warnings.simplefilter("ignore")
122 encoder = 'vitl' # can also be 'vitb' or 'vitl'
123 depth_model = DepthAnything.from_pretrained('LiheYoung/depth_anything_{:}14'.format(encoder)).cuda().eval()
124 dtransform = Compose([
125 Resize(
126 width=518,
127 height=518,
128 resize_target=False,
129 keep_aspect_ratio=True,
130 ensure_multiple_of=14,
131 resize_method='lower_bound',
132 image_interpolation_method=cv2.INTER_CUBIC,
133 ),
134 NormalizeImage(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
135 PrepareForNet(),
136 ])
137 return depth_model, dtransform
138
139# def invert_depth(depth_map):
140# inv = depth_map.copy()

Callers

nothing calls this directly

Calls 1

ResizeClass · 0.85

Tested by

no test coverage detected