JonathanGLi commited on
Commit
bf92199
·
verified ·
1 Parent(s): b2c340d
Files changed (4) hide show
  1. LICENSE +7 -0
  2. README.md +72 -3
  3. manifest.json +66 -0
  4. scafie_0815.zip +3 -0
LICENSE ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ RELEASE PENDING — NO LICENSE GRANT APPLIED YET
2
+
3
+ Proposed license: Creative Commons Attribution 4.0 International (CC BY 4.0)
4
+ https://creativecommons.org/licenses/by/4.0/
5
+ Legal terms: https://creativecommons.org/licenses/by/4.0/legalcode
6
+
7
+ Shibo Zhao recommended CC BY 4.0.
README.md CHANGED
@@ -1,3 +1,72 @@
1
- ---
2
- license: cc-by-4.0
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # SuperMap CMU Gates and Scafie pointwise annotations
2
+
3
+ This package provides source point clouds and pointwise semantic annotations for the SuperMap benchmark. It is a preliminary annotation release. Bounding-box GT, generation scripts, prediction conversion, and evaluation code are not included.
4
+
5
+ ## Downloads and sequences
6
+
7
+ | Archive | Sequences |
8
+ |---|---|
9
+ | `gates_0511.zip` | `static_first_loop`, `change_loop1`, `change_loop2` |
10
+ | `scafie_0815.zip` | `indoor_outdoor` |
11
+
12
+ These names identify sequences, not verified train/validation/test partitions. The stated six changed object instances have not been independently identified in these files.
13
+
14
+ Each sequence directory contains `point_cloud.txt`, `point_cloud.zlib`, and `label_map.json`. ZIPs have been repackaged to remove internal S3 references and annotation workflow metadata. `manifest.json` records archive sizes, SHA-256 checksums, and member lists.
15
+
16
+ ## Annotation format
17
+
18
+ Point clouds have one XYZRGB point per text line. Coordinates, units, and reference frames require owner confirmation. Labels are zlib-compressed unsigned bytes, with one label per point in the same order as the text rows. Preserve the source point order. Each sequence has its own class-ID mapping; class IDs differ between static and change annotations. Value `255` is observed frequently, but its meaning has not been confirmed. It must not be silently treated as a scored class or a confirmed ignore value.
19
+
20
+ `label_map.json` is normalized from the supplied mapping text. Mappings are provisional and should be checked with the annotation owner before quantitative evaluation. Scafie source notes state that one ladder/staircase and one fire extinguisher were labeled under TRASHCAN; these category exceptions are retained here as a known limitation.
21
+
22
+ ## Load and visualize
23
+
24
+ Decode a label vector with Python's `zlib` module:
25
+
26
+ ```python
27
+ import zlib
28
+ from pathlib import Path
29
+ labels = zlib.decompress(Path("point_cloud.zlib").read_bytes())
30
+ # Each byte corresponds to the point at the same row index in point_cloud.txt.
31
+ ```
32
+
33
+ The Gates archive includes the original `scripts/visualize_worker_response.py`, requiring Python and NumPy:
34
+
35
+ ```bash
36
+ python gates_0511/scripts/visualize_worker_response.py \
37
+ gates_0511/static_first_loop/point_cloud.txt \
38
+ gates_0511/static_first_loop/point_cloud.zlib output.ply
39
+ ```
40
+
41
+ This utility colors ID 0 red and all other points white. Its docstring calls ID 0 “OBSTACLE NOISE”; that description does not match the supplied sequence class maps. Use the output only as an ID-0 inspection visualization, not a complete semantic visualization. Its PLY output contains colored source points, not bounding boxes.
42
+
43
+ ## Bounding boxes, predictions, and evaluation
44
+
45
+ The referenced `generate_bbox_gt.py` and `generate_bbox3d_json.py` were not supplied, and no GT box `.npz` or `.ply` artifacts are available in this package. No validated bbox schema or generation command is claimed.
46
+
47
+ Team context indicates that SuperMap predictions can be exported as JSON and converted into predicted 3D boxes. The export schema, converter, coordinate transforms, matching rules, thresholds, and evaluator are not included. Obtain these from the benchmark maintainers before running or reporting bbox evaluation. A future release should specify box centers/dimensions/orientations, class and instance IDs, units, coordinate frames, and metrics.
48
+
49
+ ## Source data attribution
50
+
51
+ The associated paper authors are Shibo Zhao, Guofei Chen, Honghao Zhu, Zhiheng Li, Changwei Yao, Nader Zantout, Seungchan Kim, Wenshan Wang, Ji Zhang, and Sebastian Scherer.
52
+
53
+ The associated paper is [SuperMap: A Spatio-Temporal SLAM System for Visual-Language Navigation](https://arxiv.org/abs/2608.22896), arXiv:2608.22896 (2026). The arXiv record lists Proceedings of Robotics: Science and Systems (RSS 2026) as the journal reference. Paper authorship does not by itself identify the dataset copyright holder or release authority. Confirm any additional dataset contributors and the preferred dataset attribution before publication.
54
+
55
+ Please cite the associated paper when using this benchmark:
56
+
57
+ ```bibtex
58
+ @misc{zhao2026supermap,
59
+ title={SuperMap: A Spatio-Temporal SLAM System for Visual-Language Navigation},
60
+ author={Shibo Zhao and Guofei Chen and Honghao Zhu and Zhiheng Li and Changwei Yao and Nader Zantout and Seungchan Kim and Wenshan Wang and Ji Zhang and Sebastian Scherer},
61
+ year={2026},
62
+ eprint={2608.22896},
63
+ archivePrefix={arXiv},
64
+ primaryClass={cs.RO},
65
+ doi={10.48550/arXiv.2608.22896},
66
+ url={https://arxiv.org/abs/2608.22896}
67
+ }
68
+ ```
69
+
70
+ ## License and release status
71
+
72
+ CC BY 4.0 is the proposed dataset license, based on Shibo Zhao's recommendation.
manifest.json ADDED
@@ -0,0 +1,66 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "gates_0511.zip": {
3
+ "bytes": 354891227,
4
+ "sha256": "7134cc9dc373719eb30e4a047b145179eface0663b5a31c3807d5a8215d4bc4b",
5
+ "members": [
6
+ {
7
+ "path": "gates_0511/static_first_loop/point_cloud.txt",
8
+ "bytes": 313896906
9
+ },
10
+ {
11
+ "path": "gates_0511/static_first_loop/point_cloud.zlib",
12
+ "bytes": 221642
13
+ },
14
+ {
15
+ "path": "gates_0511/static_first_loop/label_map.json",
16
+ "bytes": 719
17
+ },
18
+ {
19
+ "path": "gates_0511/change_loop1/point_cloud.txt",
20
+ "bytes": 313896906
21
+ },
22
+ {
23
+ "path": "gates_0511/change_loop1/point_cloud.zlib",
24
+ "bytes": 23550
25
+ },
26
+ {
27
+ "path": "gates_0511/change_loop1/label_map.json",
28
+ "bytes": 455
29
+ },
30
+ {
31
+ "path": "gates_0511/change_loop2/point_cloud.txt",
32
+ "bytes": 341502570
33
+ },
34
+ {
35
+ "path": "gates_0511/change_loop2/point_cloud.zlib",
36
+ "bytes": 14261
37
+ },
38
+ {
39
+ "path": "gates_0511/change_loop2/label_map.json",
40
+ "bytes": 455
41
+ },
42
+ {
43
+ "path": "gates_0511/scripts/visualize_worker_response.py",
44
+ "bytes": 2715
45
+ }
46
+ ]
47
+ },
48
+ "scafie_0815.zip": {
49
+ "bytes": 6599876,
50
+ "sha256": "57c557c2627a5f8b52d27f96683fda488d54c111d4eb60ca85e8c830bc00e038",
51
+ "members": [
52
+ {
53
+ "path": "scafie_0815/indoor_outdoor/point_cloud.txt",
54
+ "bytes": 49422205
55
+ },
56
+ {
57
+ "path": "scafie_0815/indoor_outdoor/point_cloud.zlib",
58
+ "bytes": 38638
59
+ },
60
+ {
61
+ "path": "scafie_0815/indoor_outdoor/label_map.json",
62
+ "bytes": 802
63
+ }
64
+ ]
65
+ }
66
+ }
scafie_0815.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:57c557c2627a5f8b52d27f96683fda488d54c111d4eb60ca85e8c830bc00e038
3
+ size 6599876