Xuban commited on
Commit
a1e0cde
·
verified ·
1 Parent(s): 80311b4

Initial upload: ZipDepth b (MIT, metadata wrap of upstream checkpoint)

Browse files
Files changed (4) hide show
  1. LICENSE +21 -0
  2. LibreZipDepthb-depth.pt +3 -0
  3. NOTICE +12 -0
  4. README.md +57 -0
LICENSE ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ MIT License
2
+
3
+ Copyright (c) 2026 Fabio Tosi
4
+
5
+ Permission is hereby granted, free of charge, to any person obtaining a copy
6
+ of this software and associated documentation files (the "Software"), to deal
7
+ in the Software without restriction, including without limitation the rights
8
+ to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9
+ copies of the Software, and to permit persons to whom the Software is
10
+ furnished to do so, subject to the following conditions:
11
+
12
+ The above copyright notice and this permission notice shall be included in all
13
+ copies or substantial portions of the Software.
14
+
15
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16
+ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17
+ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18
+ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19
+ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21
+ SOFTWARE.
LibreZipDepthb-depth.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:069ebb29c136b3e4b3613fd73dadf1f83d61fa4ee4312fc601b923f12f34fe67
3
+ size 27302611
NOTICE ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ LibreZipDepth weights
2
+ ---------------------
3
+
4
+ This product contains weights derived from ZipDepth
5
+ (https://github.com/fabiotosi92/ZipDepth).
6
+ Copyright (c) 2026 Fabio Tosi.
7
+ Licensed under the MIT License.
8
+
9
+ The weights were trained by the upstream authors via knowledge distillation
10
+ from Depth Anything V2 Large pseudo-labels (~14M images across 17 public
11
+ datasets). The MIT license on these student weights is the upstream authors'
12
+ published position; this notice records the distillation lineage.
README.md ADDED
@@ -0,0 +1,57 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ library_name: libreyolo
4
+ pipeline_tag: depth-estimation
5
+ tags:
6
+ - depth-estimation
7
+ - monocular-depth
8
+ - zipdepth
9
+ ---
10
+
11
+ # LibreZipDepthb-depth
12
+
13
+ ZipDepth base: a ~6M-parameter reparameterizable CNN (RepVGG encoder, SPPF +
14
+ cross-scale neck, FPN decoder with convex upsampling) for zero-shot relative
15
+ monocular depth estimation, repackaged for LibreYOLO. This is the standard
16
+ GPU/CPU checkpoint; the sibling repo
17
+ [LibreZipDepthbnpu-depth](https://huggingface.co/LibreYOLO/LibreZipDepthbnpu-depth)
18
+ carries the separately trained unfold-free decoder for NPU/edge compilers.
19
+
20
+ ## Usage
21
+
22
+ ```python
23
+ from libreyolo import LibreYOLO
24
+
25
+ model = LibreYOLO("LibreZipDepthb-depth.pt") # auto-downloads from this repo
26
+ results = model.predict("image.jpg", save=True)
27
+ depth = results[0].depth_map.data # (H, W) relative inverse depth
28
+ ```
29
+
30
+ Outputs follow LibreYOLO's depth task contract: a dense float map on the
31
+ original image canvas, higher values mean closer to the camera, no metric unit.
32
+
33
+ ## Source
34
+
35
+ Derived from [fabiotosi92/ZipDepth](https://github.com/fabiotosi92/ZipDepth)
36
+ (`checkpoints/zipdepth_base.pth`) at commit
37
+ `6b96f4d205f8a2e5377e81c1b74cc99a47f6693a` (ECCV 2026, University of Bologna).
38
+ Copyright (c) 2026 Fabio Tosi. Licensed under the MIT License.
39
+
40
+ ## Modifications
41
+
42
+ LibreYOLO checkpoint-schema metadata wrap only. Learned parameters are
43
+ byte-identical to upstream. See `weights/convert_zipdepth_weights.py` in the
44
+ [LibreYOLO source repository](https://github.com/LibreYOLO/libreyolo).
45
+
46
+ ## Training-data provenance
47
+
48
+ Upstream trained these weights by distilling pseudo-labels produced by Depth
49
+ Anything V2 Large (a CC-BY-NC-4.0 checkpoint) over ~14M images from 17 public
50
+ datasets. The MIT grant on the resulting student weights is the upstream
51
+ authors' published licensing position; the lineage is documented here for
52
+ transparency.
53
+
54
+ ## License
55
+
56
+ MIT License. See the [`LICENSE`](./LICENSE) and [`NOTICE`](./NOTICE) files in
57
+ this repository.