| from .dpt import DepthAnythingV2 |
| import torch |
| import os |
|
|
| model_configs = { |
| 'vits': {'encoder': 'vits', 'features': 64, 'out_channels': [48, 96, 192, 384]}, |
| 'vitb': {'encoder': 'vitb', 'features': 128, 'out_channels': [96, 192, 384, 768]}, |
| 'vitl': {'encoder': 'vitl', 'features': 256, 'out_channels': [256, 512, 1024, 1024]}, |
| 'vitg': {'encoder': 'vitg', 'features': 384, 'out_channels': [1536, 1536, 1536, 1536]} |
| } |
|
|
| def build_backbone(depth_size='vitb', encoder_cond_dim=-1): |
| return DepthAnythingV2(**model_configs[depth_size], encoder_cond_dim=encoder_cond_dim) |
