File size: 2,186 Bytes
9efecd6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
UniQuery4R
Copyright 2026 Kosmo Research and the UniQuery4R authors.

This product includes software developed by third parties:

--------------------------------------------------------------------------------
VGGT (Visual Geometry Grounded Transformer)
  Source: https://github.com/facebookresearch/vggt
  Copyright (c) Meta Platforms, Inc. and affiliates. All rights reserved.
  License: VGGT License (see licenses/VGGT-LICENSE.txt), which is a
  non-commercial research license derived from the CC BY-NC 4.0 license.

  The following modules are adapted from VGGT:
    - uniquery4r/models/heads/camera_head.py
    - uniquery4r/models/heads/head_act.py
    - uniquery4r/models/heads/mlp_head.py
    - uniquery4r/models/layers/*
    - uniquery4r/utils/pose_enc.py
    - uniquery4r/utils/rotation.py

  NOTE: As required by the VGGT license, these files are distributed under the
  VGGT License terms and must not be used for commercial purposes. All other
  original UniQuery4R code is released under the Apache License 2.0 (see
  LICENSE). Redistributions of this project (or derivative works) must retain
  this NOTICE, the licenses/VGGT-LICENSE.txt file, and the copyright headers
  in the adapted source files.

--------------------------------------------------------------------------------
Depth Anything 3 (DA3)
  Source: https://github.com/ByteDance-Seed/Depth-Anything-3
  The ViT backbone configuration reference used by the 4D encoder follows the
  DA3 ViT-g model (embed_dim=1536, out_layers=[19, 27, 33, 39], alt_start=13).

--------------------------------------------------------------------------------
DINOv2
  Source: https://github.com/facebookresearch/dinov2
  Copyright (c) Meta Platforms, Inc. and affiliates. All rights reserved.
  License: Apache License 2.0.
  The ViT encoder implementation in uniquery4r/models/encoder.py is adapted
  from DINOv2 / the DA3 ViT backbone.

--------------------------------------------------------------------------------
Third-party Python packages used at runtime (torch, torchvision, numpy, einops,
trimesh, matplotlib, scipy, opencv-python, pillow, viser, gradio,
huggingface_hub) remain under their respective licenses.