cubec commited on
Commit
691052d
·
verified ·
1 Parent(s): 33a0dd1

Document mandatory and optional dataset files

Browse files
README.md CHANGED
@@ -15,6 +15,31 @@ This release supports SS, generative SLAT segmentation, and 32³ SLAT shape trai
15
  The default transfer is **166.32 GB**, including both training datasets and object shape targets.
16
  Default downloads exclude optional and deprecated folders.
17
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
18
  ## Restore tools
19
 
20
  Use the project checkout at Git revision `aac7b493e1e5e10512857a2582233c99fedda9c5`.
@@ -71,18 +96,22 @@ SS and segmentation retain those views.
71
  The export omits combined scene latents because the current shape objective uses object targets.
72
  Model weights and image-feature caches remain separate dependencies.
73
 
74
- ## Optional components
 
 
 
 
75
 
76
- | Component | Purpose | Transfer, GB |
77
  | --- | --- | ---: |
78
- | `evaluation/ycbv` | YCB-V mesh evaluation | 1.670 |
79
- | `geometry/3dfront` | Mesh scoring and new target encoding | 6.252 |
80
- | `geometry/sage` | Mesh scoring and new target encoding | 303.693 |
81
- | `held_out/3dfront` | Remaining test observations | 22.440 |
82
- | `held_out/foundationpose` | Remaining test observations | 6.825 |
83
- | `held_out/sage` | Remaining test observations | 8.183 |
84
- | `occlusion/3dfront` | GT relations for SS object attention | 0.240 |
85
- | `occlusion/sage` | GT relations for SS object attention | 0.103 |
86
 
87
  Restore observations before geometry, occlusion, or test overlays.
88
  YCB-V evaluation includes its own observations and geometry.
 
15
  The default transfer is **166.32 GB**, including both training datasets and object shape targets.
16
  Default downloads exclude optional and deprecated folders.
17
 
18
+ <!-- partprobe3d-required-files:start -->
19
+ ## Mandatory files
20
+
21
+ These paths refer to the Hugging Face repository root.
22
+ Files are mandatory for the task specified in each row.
23
+ The default command restores training files.
24
+ The benchmark also requires the separate download with `--dataset evaluation`.
25
+
26
+ | File or directory | Required task | Contents |
27
+ | --- | --- | --- |
28
+ | `training/index.json` | Training restoration | Component paths, sizes, monitoring counts, and source identities |
29
+ | `training/observations/<dataset>/` | Training on the selected dataset | Archive index, sample NPZ files, calibration, object mappings, and scene voxel targets |
30
+ | `training/ss/<dataset>/` | Direct SS training after target publication | Archive index, native SS latents, and codec provenance |
31
+ | `training/shape/32/<dataset>/` | SLAT shape training where targets are available | Archive index and object shape targets with explicit exclusions |
32
+ | `evaluation/index.json` | Benchmark restoration | Archive list and SHA-256 checksums |
33
+ | `evaluation/metadata.tar` | Benchmark evaluation | Portable manifests, source selections, and dataset documentation |
34
+ | `evaluation/picasso-*.tar` | Benchmark evaluation | All 100 PICASSO images, masks, previews, and GT meshes |
35
+ | `evaluation/messykitchen-*.tar` | Benchmark evaluation | All 1,503 MessyKitchens images, masks, previews, and 100 shared GT meshes |
36
+ | `evaluation/3dfront-*.tar` | Benchmark evaluation | The selected 1,000 3D-FRONT test images, masks, previews, and GT meshes |
37
+
38
+ Restore every archive listed in the selected component's `index.json`.
39
+ The evaluation bundle has 26 mandatory archives and requires about 23.4 GB.
40
+ Older checkouts also require the corresponding restore patch under `code/`.
41
+ <!-- partprobe3d-required-files:end -->
42
+
43
  ## Restore tools
44
 
45
  Use the project checkout at Git revision `aac7b493e1e5e10512857a2582233c99fedda9c5`.
 
96
  The export omits combined scene latents because the current shape objective uses object targets.
97
  Model weights and image-feature caches remain separate dependencies.
98
 
99
+ ## Optional files
100
+
101
+ These files support additional datasets, test views, and target generation.
102
+ The mandatory evaluation bundle already includes its RGB images, masks, and GT meshes.
103
+ Some SS configurations require the matching optional occlusion component.
104
 
105
+ | File or directory | Purpose | Transfer, GB |
106
  | --- | --- | ---: |
107
+ | `optional/evaluation/ycbv/` | YCB-V mesh evaluation | 1.670 |
108
+ | `optional/geometry/3dfront/` | Mesh scoring and new target encoding | 6.252 |
109
+ | `optional/geometry/sage/` | Mesh scoring and new target encoding | 303.693 |
110
+ | `optional/held_out/3dfront/` | Remaining test observations | 22.440 |
111
+ | `optional/held_out/foundationpose/` | Remaining test observations | 6.825 |
112
+ | `optional/held_out/sage/` | Remaining test observations | 8.183 |
113
+ | `optional/occlusion/3dfront/` | GT relations for SS object attention | 0.240 |
114
+ | `optional/occlusion/sage/` | GT relations for SS object attention | 0.103 |
115
 
116
  Restore observations before geometry, occlusion, or test overlays.
117
  YCB-V evaluation includes its own observations and geometry.
evaluation/README.md CHANGED
@@ -17,6 +17,38 @@ The selected views cover 635 rooms and retain their test assignments.
17
  Each MessyKitchens scene has one GT export, shared across its views through relative manifest paths.
18
  All archive members contain regular files.
19
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
20
  ## Restore
21
 
22
  Use the PartProbe3D checkout at Git revision `290027fe9f281a21db978a48da21bb9be90e5517`.
 
17
  Each MessyKitchens scene has one GT export, shared across its views through relative manifest paths.
18
  All archive members contain regular files.
19
 
20
+ ## Mandatory files
21
+
22
+ These paths refer to the Hugging Face repository root.
23
+ The complete benchmark requires `evaluation/index.json` and all 26 archives listed below.
24
+
25
+ | File or files | Count | Contents |
26
+ | --- | ---: | --- |
27
+ | `evaluation/index.json` | 1 | Archive inventory and SHA-256 checksums |
28
+ | `evaluation/metadata.tar` | 1 | Portable manifests, original selections, and dataset documentation |
29
+ | `evaluation/picasso-00000.tar` | 1 | All 100 PICASSO images, masks, previews, and GT meshes |
30
+ | `evaluation/messykitchen-00001.tar` through `messykitchen-00022.tar` | 22 | All 1,503 MessyKitchens images, masks, previews, and 100 shared GT meshes |
31
+ | `evaluation/3dfront-00023.tar` and `3dfront-00024.tar` | 2 | The selected 1,000 3D-FRONT test images, masks, previews, and GT meshes |
32
+
33
+ The downloader restores every archive and checks its checksum.
34
+ Each view requires its RGB image, instance mask, and referenced GT mesh.
35
+ Older project checkouts also require `code/evaluation.patch` before restoration.
36
+
37
+ ## Optional files
38
+
39
+ These separate Hub components are optional for this benchmark:
40
+
41
+ | File or directory | Purpose |
42
+ | --- | --- |
43
+ | `optional/held_out/<dataset>/` | Remaining test observations beyond the training monitors |
44
+ | `optional/geometry/<dataset>/` | Source geometry for new targets or other evaluation sets |
45
+ | `optional/occlusion/<dataset>/` | GT relations for applicable SS configurations |
46
+ | `optional/evaluation/ycbv/` | Additional YCB-V evaluation data |
47
+
48
+ The mandatory bundle already includes the GT meshes for all 2,603 images.
49
+ Its restoration needs no optional component or original source dataset.
50
+ Training files under `training/` serve model training separately.
51
+
52
  ## Restore
53
 
54
  Use the PartProbe3D checkout at Git revision `290027fe9f281a21db978a48da21bb9be90e5517`.
evaluation/index.json CHANGED
@@ -31,8 +31,8 @@
31
  {
32
  "file": "metadata.tar",
33
  "bytes": 13086720,
34
- "sha256": "9e5e5ea7f7b3dd012b6f9484f14d41758e0ed5843e0cec682a25ebc0c507989c",
35
- "identity": "1d3d20302be51b87309bef221f8516ced8bb82d0a8644aae6844c0c63a8704b0",
36
  "files": 9,
37
  "views": 0
38
  },
 
31
  {
32
  "file": "metadata.tar",
33
  "bytes": 13086720,
34
+ "sha256": "d771d642708809843cf7cfca7ceb342e1b687deccbd135ee30e9b74402f0883f",
35
+ "identity": "4dac3374b520102aff127fc5e4cd6ab41da2ed03539523a05c98eb9e3374854d",
36
  "files": 9,
37
  "views": 0
38
  },
evaluation/metadata.tar CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9e5e5ea7f7b3dd012b6f9484f14d41758e0ed5843e0cec682a25ebc0c507989c
3
  size 13086720
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d771d642708809843cf7cfca7ceb342e1b687deccbd135ee30e9b74402f0883f
3
  size 13086720