GeorgeWJJ commited on
Commit
dbc1584
·
verified ·
1 Parent(s): 835f7d1

Finalize bilingual cards, mesh/conversion documentation and complete release checksums

Browse files
Files changed (4) hide show
  1. README.md +18 -5
  2. README.zh-CN.md +17 -5
  3. SHA256SUMS +31 -2
  4. assets/sampling_preview.png +3 -0
README.md CHANGED
@@ -1,4 +1,5 @@
1
  ---
 
2
  pretty_name: MCB-32 Paired Industrial Point Clouds
3
  language:
4
  - en
@@ -17,7 +18,7 @@ tags:
17
  ---
18
  # MCB-32 Paired Industrial Point Clouds
19
 
20
- [English](README.md) | [简体中文](README.zh-CN.md) · [Code and reports](https://github.com/13536309143/pointnext)
21
 
22
  ## Dataset summary
23
 
@@ -25,7 +26,11 @@ A research preparation of Mechanical Components Benchmark (MCB) assets for **who
25
 
26
  The release contains area-weighted random, greedy farthest-point and Poisson-disk-style samples at **2,048, 4,096 and 8,192 points**. Each file preserves a contiguous NumPy `float32` array of shape `(N,3)`, XYZ only. No color, normals, physical units, sensor measurements or point-wise labels are included. Labels describe component shape and function, not defects or materials.
27
 
28
- The source OBJ collection is retained locally for conversion and is not bundled into these point cloud archives.
 
 
 
 
29
 
30
  ## Files and partitions
31
 
@@ -38,6 +43,8 @@ The source OBJ collection is retained locally for conversion and is not bundled
38
  | Point cloud files | 454,779 |
39
  | NPY payload bytes | 26,137,058,688 (24.34 GiB) |
40
  | Representations | Three methods ×three point counts |
 
 
41
 
42
  Files are delivered as approximately 512 MiB **uncompressed TAR shards** to avoid hundreds of thousands of small Hub entries. Sharding preserves every NPY byte and path; TAR headers/padding add some overhead. `release_manifest.json` lists archive sizes, SHA-256 hashes, variant, original partition and file counts. `SHA256SUMS` also covers metadata and documentation. This is a file-based NumPy dataset, not a preconfigured Hugging Face `load_dataset()` table.
43
 
@@ -50,6 +57,10 @@ metadata/experiment_splits.csv.gz
50
  metadata/file_manifest.csv.gz
51
  release_manifest.json
52
  SHA256SUMS
 
 
 
 
53
  ```
54
 
55
  Each archive restores paths of the form:
@@ -74,7 +85,9 @@ python -m pip install huggingface_hub numpy
74
  python scripts/download_dataset.py --repo-id GeorgeWJJ/PointNext --output data
75
  ```
76
 
77
- For a single representation, add `--variant area_random_2048`. The full experiment suite requires all nine. Supply `--revision <dataset-commit>` to pin the exact version. The downloader verifies archive SHA-256, rejects unsafe archive paths and restores the original directory structure. It does not download the source meshes.
 
 
78
 
79
  ```python
80
  from pathlib import Path
@@ -143,7 +156,7 @@ The largest class, `screws_and_bolts`, contains 16,530 objects (32.71%). `flange
143
 
144
  Suitable for CAD surface classification, paired sampling/resolution studies and controlled synthetic corruption experiments. Inputs are ideal complete surfaces, not real scans. The release does not establish robustness to arbitrary poses, outliers, occlusion, sensor artifacts or unseen CAD sources. Labels merge several fine subtypes into broad component categories. The 32-class taxonomy and retained counts are specific to this preparation.
145
 
146
- The associated **completed** experiment trains six classifiers on `area_random_2048` only; the remaining variants are cross-representation tests on the same held-out objects. It is not a nine-training-dataset optimization study and has only one training seed. The [English report](https://github.com/13536309143/pointnext/blob/main/docs/reports/KEY6.md) and [Chinese report](https://github.com/13536309143/pointnext/blob/main/docs/reports/KEY6.zh-CN.md) provide measured results.
147
 
148
  ## Provenance, rights and citation
149
 
@@ -160,4 +173,4 @@ Original source: [stnoah1/mcb](https://github.com/stnoah1/mcb), introduced by Ki
160
  }
161
  ```
162
 
163
- Dataset preparation and experiment code: [13536309143/pointnext](https://github.com/13536309143/pointnext). This Hub repository is a derivative research distribution, not an endorsement by the original MCB or PointNeXt authors.
 
1
  ---
2
+ viewer: false
3
  pretty_name: MCB-32 Paired Industrial Point Clouds
4
  language:
5
  - en
 
18
  ---
19
  # MCB-32 Paired Industrial Point Clouds
20
 
21
+ [English](README.md) | [简体中文](README.zh-CN.md) · [Code and reports](https://github.com/13536309143/MCB-32-Industrial-Point-Cloud-Classification)
22
 
23
  ## Dataset summary
24
 
 
26
 
27
  The release contains area-weighted random, greedy farthest-point and Poisson-disk-style samples at **2,048, 4,096 and 8,192 points**. Each file preserves a contiguous NumPy `float32` array of shape `(N,3)`, XYZ only. No color, normals, physical units, sensor measurements or point-wise labels are included. Labels describe component shape and function, not defects or materials.
28
 
29
+ All **50,531 prepared source OBJ meshes** are also available as an optional, separately compressed `meshes/` collection. They preserve the same IDs, classes and partitions. Conversion sources and the Windows one-click wrapper are mirrored under `conversion/`; see [conversion instructions](https://huggingface.co/datasets/GeorgeWJJ/PointNext/blob/main/conversion/README.md).
30
+
31
+ ![Nine paired representations of one object](assets/sampling_preview.png)
32
+
33
+ The same held-out gear is shown with a fixed camera and coordinate frame. This is a data illustration, not a classification success example. The tabular dataset viewer is disabled because this release contains original NPY files in TAR shards; use the verified downloader below.
34
 
35
  ## Files and partitions
36
 
 
43
  | Point cloud files | 454,779 |
44
  | NPY payload bytes | 26,137,058,688 (24.34 GiB) |
45
  | Representations | Three methods ×three point counts |
46
+ | Optional prepared OBJ meshes | 50,531 files; 8,909,142,490 bytes (8.30 GiB) |
47
+ | Mesh download | 18 gzip TAR shards; 2,654,025,959 bytes (2.47 GiB) |
48
 
49
  Files are delivered as approximately 512 MiB **uncompressed TAR shards** to avoid hundreds of thousands of small Hub entries. Sharding preserves every NPY byte and path; TAR headers/padding add some overhead. `release_manifest.json` lists archive sizes, SHA-256 hashes, variant, original partition and file counts. `SHA256SUMS` also covers metadata and documentation. This is a file-based NumPy dataset, not a preconfigured Hugging Face `load_dataset()` table.
50
 
 
57
  metadata/file_manifest.csv.gz
58
  release_manifest.json
59
  SHA256SUMS
60
+ meshes/<train-or-test>-<shard>.tar.gz # optional original OBJ meshes
61
+ mesh_release_manifest.json
62
+ metadata/mesh_file_manifest.csv.gz
63
+ conversion/ # converter, C++ source, Windows DLL and BAT
64
  ```
65
 
66
  Each archive restores paths of the form:
 
85
  python scripts/download_dataset.py --repo-id GeorgeWJJ/PointNext --output data
86
  ```
87
 
88
+ The default Hub download cache retains the TAR shards in addition to the extracted NPY files; allow approximately 55 GiB of free disk for the full download/extraction, plus space for training outputs.
89
+
90
+ For a single representation, add `--variant area_random_2048`. The full experiment suite requires all nine. Supply `--revision <dataset-commit>` to pin the exact version. The downloader verifies archive SHA-256, rejects unsafe archive paths and restores the original directory structure. The default downloads only point clouds. Add `--include-meshes` to also restore the original `MCB_merged/`, or use `--meshes-only` to download just the OBJ collection. Mesh shards use lossless gzip compression; `mesh_release_manifest.json` and `metadata/mesh_file_manifest.csv.gz` provide archive and individual OBJ hashes. Allow extra cache and extraction space for meshes (about 8.3 GiB extracted, plus compressed downloads).
91
 
92
  ```python
93
  from pathlib import Path
 
156
 
157
  Suitable for CAD surface classification, paired sampling/resolution studies and controlled synthetic corruption experiments. Inputs are ideal complete surfaces, not real scans. The release does not establish robustness to arbitrary poses, outliers, occlusion, sensor artifacts or unseen CAD sources. Labels merge several fine subtypes into broad component categories. The 32-class taxonomy and retained counts are specific to this preparation.
158
 
159
+ The associated **completed** experiment trains six classifiers on `area_random_2048` only; the remaining variants are cross-representation tests on the same held-out objects. It is not a nine-training-dataset optimization study and has only one training seed. The [English report](https://github.com/13536309143/MCB-32-Industrial-Point-Cloud-Classification/blob/main/docs/reports/KEY6.md) and [Chinese report](https://github.com/13536309143/MCB-32-Industrial-Point-Cloud-Classification/blob/main/docs/reports/KEY6.zh-CN.md) provide measured results.
160
 
161
  ## Provenance, rights and citation
162
 
 
173
  }
174
  ```
175
 
176
+ Dataset preparation and experiment code: [13536309143/MCB-32-Industrial-Point-Cloud-Classification](https://github.com/13536309143/MCB-32-Industrial-Point-Cloud-Classification). This Hub repository is a derivative research distribution, not an endorsement by the original MCB or PointNeXt authors.
README.zh-CN.md CHANGED
@@ -1,6 +1,6 @@
1
  # MCB-32 工业零部件配对点云
2
 
3
- [English](README.md) | [简体中文](README.zh-CN.md) · [代码与报告](https://github.com/13536309143/pointnext)
4
 
5
  ## 概况
6
 
@@ -8,12 +8,18 @@
8
 
9
  三种方法为按面积随机采样、贪心最远点采样(FPS)、泊松盘式样本消除,点数分别为 **2,048、4,096、8,192**。所有文件均为连续存储的 NumPy `float32`,形状 `(N,3)`,只含 XYZ,不含法线、颜色、物理尺寸、传感器测量或逐点标签。类别表达机械零部件类型,同时涉及形状和用途,不是材质或缺陷标签。
10
 
11
- 本次分发九套点云;原始合并 OBJ 留在本地用于重新转换,不包含在点云归档中。
 
 
 
 
12
 
13
  ## 数量与文件
14
 
15
  每套原始 train/test 为 **40,424 / 10,107**;对应已完成实验的 train/val/test 为 **36,380 / 4,044 / 10,107**。共 454,779 个 NPY,逻辑字节数 **26,137,058,688(24.34 GiB)**。
16
 
 
 
17
  为了避免数十万个仓库条目,使用约 512 MiB 的无压缩 TAR 分片,保留原始 NPY 的字节和相对路径;TAR 头与补齐会增加一定体积。`release_manifest.json` 记录每片路径、大小、SHA-256、版本、原始划分及文件数,`SHA256SUMS` 另覆盖元数据与说明文档。本数据是 NumPy 文件集合,不是已经配置好的 `load_dataset()` 表格。
18
 
19
  ```text
@@ -25,6 +31,10 @@ metadata/experiment_splits.csv.gz
25
  metadata/file_manifest.csv.gz
26
  release_manifest.json
27
  SHA256SUMS
 
 
 
 
28
  ```
29
 
30
  归档内部保留:
@@ -53,7 +63,9 @@ python -m pip install huggingface_hub numpy
53
  python scripts/download_dataset.py --repo-id GeorgeWJJ/PointNext --output data
54
  ```
55
 
56
- 只下载一个版本可增加 `--variant area_random_2048`;完整实验需要九套。为复现准确版本可添加 `--revision <数据集提交号>`。程序核对分片 SHA-256、拒绝越界或链接类归档路径,恢复原始 NPY 目录;不下载原始网格。
 
 
57
 
58
  ```python
59
  from pathlib import Path
@@ -121,7 +133,7 @@ label = p.parent.name
121
 
122
  适合完整 CAD 表面的分类、配对采样/分辨率研究和人工扰动测试;不是实际扫描数据。尚不能代表真实设备噪声、任意姿态、离群点、遮挡或新 CAD 来源的泛化。32 类和保留模型数量属于本整理版。
123
 
124
- 关联的**已完成实验只在面积随机 2,048 点上训练六个网络**,另外八套仅对相同留出对象做跨表示测试;仅一个训练种子。不能据此判断九种训练数据的最优方案。详细结果见 [中文报告](https://github.com/13536309143/pointnext/blob/main/docs/reports/KEY6.zh-CN.md)与[英文报告](https://github.com/13536309143/pointnext/blob/main/docs/reports/KEY6.md)。
125
 
126
  ## 来源、权利与引用
127
 
@@ -140,4 +152,4 @@ label = p.parent.name
140
  }
141
  ```
142
 
143
- 整理与实验���码:[13536309143/pointnext](https://github.com/13536309143/pointnext)。本发布不代表原 MCB 或 PointNeXt 作者背书。
 
1
  # MCB-32 工业零部件配对点云
2
 
3
+ [English](README.md) | [简体中文](README.zh-CN.md) · [代码与报告](https://github.com/13536309143/MCB-32-Industrial-Point-Cloud-Classification)
4
 
5
  ## 概况
6
 
 
8
 
9
  三种方法为按面积随机采样、贪心最远点采样(FPS)、泊松盘式样本消除,点数分别为 **2,048、4,096、8,192**。所有文件均为连续存储的 NumPy `float32`,形状 `(N,3)`,只含 XYZ,不含法线、颜色、物理尺寸、传感器测量或逐点标签。类别表达机械零部件类型,同时涉及形状和用途,不是材质或缺陷标签。
10
 
11
+ 本次分发九套点云,并通过独立的 `meshes/` 压缩分片提供全部 **50,531 个整理后的源 OBJ 网格**,其模型 ID、类别和原始划分与点云一一对应。`conversion/` 同时附有转换脚本、C++ 采样源码、Windows DLL 和一键 BAT,参见[转换说明](https://huggingface.co/datasets/GeorgeWJJ/PointNext/blob/main/conversion/README.md)。
12
+
13
+ ![同一对象的九种配对点云](assets/sampling_preview.png)
14
+
15
+ 同一个留出 gear,使用固定视角与坐标系,仅展示数据,不是识别成功案例。本发布为原始 NPY 的 TAR 分片,已关闭不适用的表格预览;请按下面的方法下载与校验。
16
 
17
  ## 数量与文件
18
 
19
  每套原始 train/test 为 **40,424 / 10,107**;对应已完成实验的 train/val/test 为 **36,380 / 4,044 / 10,107**。共 454,779 个 NPY,逻辑字节数 **26,137,058,688(24.34 GiB)**。
20
 
21
+ 可选 OBJ 集合共 **50,531 个文件,8,909,142,490 字节(8.30 GiB)**;无损压缩为 **18 个 TAR.GZ 分片,2,654,025,959 字节(2.47 GiB)**。
22
+
23
  为了避免数十万个仓库条目,使用约 512 MiB 的无压缩 TAR 分片,保留原始 NPY 的字节和相对路径;TAR 头与补齐会增加一定体积。`release_manifest.json` 记录每片路径、大小、SHA-256、版本、原始划分及文件数,`SHA256SUMS` 另覆盖元数据与说明文档。本数据是 NumPy 文件集合,不是已经配置好的 `load_dataset()` 表格。
24
 
25
  ```text
 
31
  metadata/file_manifest.csv.gz
32
  release_manifest.json
33
  SHA256SUMS
34
+ meshes/<train-or-test>-<shard>.tar.gz
35
+ mesh_release_manifest.json
36
+ metadata/mesh_file_manifest.csv.gz
37
+ conversion/
38
  ```
39
 
40
  归档内部保留:
 
63
  python scripts/download_dataset.py --repo-id GeorgeWJJ/PointNext --output data
64
  ```
65
 
66
+ 默认 Hugging Face 缓存会在解压后保留 TAR。完整下载与解压建议预留约 55 GiB 空闲磁盘,训练输出另计。
67
+
68
+ 只下载一个版本可增加 `--variant area_random_2048`;完整实验需要九套。为复现准确版本可添加 `--revision <数据集提交号>`。程序核对分片 SHA-256、拒绝越界或链接类归档路径,恢复原始 NPY 目录;默认只下载点云。添加 `--include-meshes` 可同时恢复 `MCB_merged/`;使用 `--meshes-only` 则只下载 OBJ。网格使用无损 gzip 压缩,`mesh_release_manifest.json` 记录分片哈希,`metadata/mesh_file_manifest.csv.gz` 记录每个 OBJ 的哈希。网格解压另需约 8.3 GiB,加上压缩分片缓存空间。
69
 
70
  ```python
71
  from pathlib import Path
 
133
 
134
  适合完整 CAD 表面的分类、配对采样/分辨率研究和人工扰动测试;不是实际扫描数据。尚不能代表真实设备噪声、任意姿态、离群点、遮挡或新 CAD 来源的泛化。32 类和保留模型数量属于本整理版。
135
 
136
+ 关联的**已完成实验只在面积随机 2,048 点上训练六个网络**,另外八套仅对相同留出对象做跨表示测试;仅一个训练种子。不能据此判断九种训练数据的最优方案。详细结果见 [中文报告](https://github.com/13536309143/MCB-32-Industrial-Point-Cloud-Classification/blob/main/docs/reports/KEY6.zh-CN.md)与[英文报告](https://github.com/13536309143/MCB-32-Industrial-Point-Cloud-Classification/blob/main/docs/reports/KEY6.md)。
137
 
138
  ## 来源、权利与引用
139
 
 
152
  }
153
  ```
154
 
155
+ 整理与实验代码:[13536309143/MCB-32-Industrial-Point-Cloud-Classification](https://github.com/13536309143/MCB-32-Industrial-Point-Cloud-Classification)。本发布不代表原 MCB 或 PointNeXt 作者背书。
SHA256SUMS CHANGED
@@ -1,5 +1,5 @@
1
- 2bba153d026247acf5b8f514b9ec75340c9021c9f216149fd2d375fca89d6da3 README.md
2
- 6609fa232f0d8e3d20d890839c49b43d0f8c4acccc6d6fb9f365a04fdec4cd8c README.zh-CN.md
3
  20510089f1b2d54458060a39e63cf370c285feed38cd29db910e70bfb942dfc3 THIRD_PARTY_NOTICES.md
4
  c2c4f1dac2a88a5d1418c67151f535bba35d84acfc044e1114e636cdb1f292ac archives/area_random_2048/test-00000.tar
5
  1f3b29f6573400be900f6386c8e557ec5f98d110ae65911f295b7b9f7c1b6737 archives/area_random_2048/train-00000.tar
@@ -55,8 +55,37 @@ f28a9ae96dcb925d6b41dbec73cd04d85243f7e5312ff45396c38e5955176783 archives/poiss
55
  4509d6812bef8642e721d0518fb9a4c5dce27913e9b3e44f2f1bb7a64c97b223 archives/poisson_disk_8192/train-00005.tar
56
  d2b55874fa2615674b3270f73ca37e6adcf09b027e1e7b6770df588919d805c9 archives/poisson_disk_8192/train-00006.tar
57
  40fc8fe58c0696043f4a26de28ba9f597e954850932a5034bd676ffe3dd53003 archives/poisson_disk_8192/train-00007.tar
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
58
  7220be7260acb63a10ce6b0a865d82bd7ba1f370771e7d5f6b2fe77d3179755c metadata/class_counts.json
59
  832ee93e3d419c3cbaa5bea31a1aeb65955a713b61615de17b18b87a273b7525 metadata/class_to_idx.json
60
  1143776387b86e6682a9355d87654622a883dc567fc40fa065bad78e3f0e5e6e metadata/experiment_splits.csv.gz
61
  0f2391defd76efb762f1e13d9ab9b10a2eaf60bfc20c1b393f59ccb36bdd891e metadata/file_manifest.csv.gz
 
62
  bb1c9d96782f9a5614271fcb187aa256c48a829e1ff3c1c37ada9dab8052bd55 release_manifest.json
 
1
+ ca3fb861c6ef4a9d4d75ca69b91af45a0ad93031f2ae3611a25b16367313cd44 README.md
2
+ 3329ba5fdcdf9244c6a43367fd91b48223c8a329a1662c00a4d607883b7b4652 README.zh-CN.md
3
  20510089f1b2d54458060a39e63cf370c285feed38cd29db910e70bfb942dfc3 THIRD_PARTY_NOTICES.md
4
  c2c4f1dac2a88a5d1418c67151f535bba35d84acfc044e1114e636cdb1f292ac archives/area_random_2048/test-00000.tar
5
  1f3b29f6573400be900f6386c8e557ec5f98d110ae65911f295b7b9f7c1b6737 archives/area_random_2048/train-00000.tar
 
55
  4509d6812bef8642e721d0518fb9a4c5dce27913e9b3e44f2f1bb7a64c97b223 archives/poisson_disk_8192/train-00005.tar
56
  d2b55874fa2615674b3270f73ca37e6adcf09b027e1e7b6770df588919d805c9 archives/poisson_disk_8192/train-00006.tar
57
  40fc8fe58c0696043f4a26de28ba9f597e954850932a5034bd676ffe3dd53003 archives/poisson_disk_8192/train-00007.tar
58
+ b647466871fccf421a4aba87dd6a941caec4970380bb9bd69bdc543d8d1da520 assets/sampling_preview.png
59
+ 0b783f4e7d0d854c64f5d924a7b1357f2605959d64e793c06ee561e6fa18b6e1 conversion/LICENSE
60
+ 4f70ebae2d60e6c885104b7a77d5f23a56954f6f99e7eb2ef6af1c519a95860f conversion/README.md
61
+ 4315bb9853ef77f183d0b5be10b6117dbf84044bb3b5b4f234451fa7f70303b3 conversion/README.zh-CN.md
62
+ ba85aaaab2c3a11dd769a5bc05898285999811aaf0275fab3f2bd0d58b784801 conversion/_pointcloud_converter/bootstrap.ps1
63
+ d66a58ff4a4e1c5a42862a54df13e5465fa0e13092d3ce5e341f457f7d7ea356 conversion/_pointcloud_converter/convert_mcb.py
64
+ 5069f976731386b3565882fef7aa5c194ea303e313263025d2d22dad8458a161 conversion/_pointcloud_converter/sampling.cpp
65
+ ea13751325f7ad28dc5eafa624230f034e8066f6635e927d58b72074c8bbd947 conversion/_pointcloud_converter/sampling_windows_x64.dll
66
+ 7cc46fcc21b8e2ef54d30f86a52086254d29edb7e6c58acb91695c8fbf5f221f conversion/convert_mcb_9sets.bat
67
+ 251e3516f3caa236776fcdf6db5f76394df9374b65c44e08d2cb6bc09117989a mesh_release_manifest.json
68
+ ba7eea5d54e794d7825c976a1f6c9787e019e0fbe27239ad551099fca2c9f7d7 meshes/test-00000.tar.gz
69
+ ad38b0057feb9fe5acfdf9ecd832f60495fa7845509b8d0bfabd99b8266bde25 meshes/test-00001.tar.gz
70
+ 3f5c0a708391b6c7fe1f169f249e84e74321cd1d6380ed32347fe43a412835e1 meshes/test-00002.tar.gz
71
+ 6643d50edc1f9e8f7423e1449f83998aac576397e5fd48a42ff18c7d6476d01b meshes/test-00003.tar.gz
72
+ d243a888f16740220e8907eac8a86161aa48b541b11a08026dfd1eac8e2f9769 meshes/train-00000.tar.gz
73
+ 97f18a1f6022964f74dc9d00639f5861fa324733f53ab7cab3e80101fe77e6f9 meshes/train-00001.tar.gz
74
+ b8c9449cd7304f75581a7d6741ed04b7f9613313906c45ced1d992924fe35085 meshes/train-00002.tar.gz
75
+ 575e0862407befc353f8bf786bd21de1b8764a2e6d66af92f2b156ff188f900f meshes/train-00003.tar.gz
76
+ 6d4b7636aa222ef6d5ca186320b12f9584e87d8f82b414c22285b3fcc1f7bc5b meshes/train-00004.tar.gz
77
+ 6935b2e4d971f4f6dd5fcdc6c3a0d5769e3a512c9917e684b3af824a54b3f85d meshes/train-00005.tar.gz
78
+ 2bafd68d073f174fa4be886cc84b356a4213cf13f8806e9fd6a2d38aa270dc18 meshes/train-00006.tar.gz
79
+ 9e4566eed035face2804d0372962d2a52c81729b7a7b59b816f77137a786dd61 meshes/train-00007.tar.gz
80
+ 16803e5a1693ffc52e2fb50a0e5c22afbbebfa89505d0b46d76513b1318d0e38 meshes/train-00008.tar.gz
81
+ 9f0fd44b4b35ee8a6061abba2142e3bae62f703dc7a69b59cfe0a3243b09315e meshes/train-00009.tar.gz
82
+ 254a3d96d67fb1afbf49e15ce89acf9475c30454fa4b900443c18477bfc4bd9c meshes/train-00010.tar.gz
83
+ 97777b2da8557ee194e93b51f51f42220b7da258cdacf59d34584ca9c823dbc4 meshes/train-00011.tar.gz
84
+ 8901edf4a88e993cd877cfcf5110aefea769f22dfeed89ce2ce4f138711650a7 meshes/train-00012.tar.gz
85
+ bb32d2d64d562668acad1ef251671f1684ea8fc76c613db75b7fccadaa11c4e2 meshes/train-00013.tar.gz
86
  7220be7260acb63a10ce6b0a865d82bd7ba1f370771e7d5f6b2fe77d3179755c metadata/class_counts.json
87
  832ee93e3d419c3cbaa5bea31a1aeb65955a713b61615de17b18b87a273b7525 metadata/class_to_idx.json
88
  1143776387b86e6682a9355d87654622a883dc567fc40fa065bad78e3f0e5e6e metadata/experiment_splits.csv.gz
89
  0f2391defd76efb762f1e13d9ab9b10a2eaf60bfc20c1b393f59ccb36bdd891e metadata/file_manifest.csv.gz
90
+ a79d6f8a97727702f4b1549fb2261913cf4560172d59b8f4e45d52701f55c253 metadata/mesh_file_manifest.csv.gz
91
  bb1c9d96782f9a5614271fcb187aa256c48a829e1ff3c1c37ada9dab8052bd55 release_manifest.json
assets/sampling_preview.png ADDED

Git LFS Details

  • SHA256: b647466871fccf421a4aba87dd6a941caec4970380bb9bd69bdc543d8d1da520
  • Pointer size: 132 Bytes
  • Size of remote file: 1.16 MB