warmhammer commited on
Commit
d7f9824
·
verified ·
1 Parent(s): 891b91a

Readme: adding installation info

Browse files
Files changed (1) hide show
  1. README.md +101 -85
README.md CHANGED
@@ -1,86 +1,102 @@
1
- ---
2
- license: cc-by-4.0
3
- ---
4
-
5
- # OSMa-Bench Dataset
6
-
7
- [![OSMa-Bench pipeline](https://be2rlab.github.io/OSMa-Bench/static/images/full_pipeline.png)](https://be2rlab.github.io/OSMa-Bench/)
8
-
9
- OSMa-Bench (Open Semantic Mapping Benchmark) dataset is a fully automatically generated dataset for evaluating the robustness of open semantic mapping and segmentation systems under varying indoor lighting conditions and robot movement dynamics. This dataset is part of [OSMa-Bench](https://be2rlab.github.io/OSMa-Bench/) pipeline.
10
-
11
- - **Homepage:** [OSMa-Bench Project Page](https://be2rlab.github.io/OSMa-Bench/)
12
-
13
- ## Dataset Summary
14
-
15
- This dataset provides simulated RGB-D and semantically annotated posed sequences for evaluation of semantic mapping and segmentation, with a particular focus on handling dynamic lighting—a critical but often overlooked factor in existing benchmarks. It also includes a collection of automatically generated question–answer pairs across multiple categories to support the evaluation of scene-graph–based reasoning, offering a task-driven measure of how well a system’s reconstructed scene captures semantic relationships between objects.
16
-
17
- The data is built upon two base datasets:
18
- - **ReplicaCAD**: 22 scenes with 4 lighting configurations and a velocity modifier.
19
- - **Habitat Matterport 3D (HM3D)**: 8 scenes with 2 lighting configurations and a velocity modifier.
20
-
21
- ## Data Configurations
22
-
23
- The dataset includes the following configurations for the ReplicaCAD and HM3D scenes:
24
-
25
- | Configuration | Description |
26
- | :--- | :--- |
27
- | `baseline` | Static, non-uniformly distributed light sources (ReplicaCAD only) |
28
- | `dynamic_lighting` | Lighting conditions change along the robot's path (ReplicaCAD only) |
29
- | `nominal_lights` | The mesh itself emits light without added light sources |
30
- | `camera_light` | An extra directed light source is attached to the camera |
31
- | `velocity` | Sequences recorded at doubled nominal velocity |
32
-
33
- ---
34
-
35
- ## Data Configurations
36
-
37
- The dataset provides structured data for each scene, suitable for tasks like 3D scene understanding, visual question answering, and robotics. Each scene contains the following components:
38
-
39
- | Component | Description | Format / Example |
40
- | ------------------------- | ------------------------------------------------------------------------------------------------- | --------------------------------------|
41
- | **RGB Images** | Standard color images captured from different camera viewpoints. | `frame000000.jpg`, ... |
42
- | **Depth Images** | Depth maps aligned with RGB images. Each pixel encodes depth in meters. | `depth000000.png`, ... |
43
- | **Semantic Masks** | Pixel-wise semantic segmentation labels. Each pixel corresponds to a semantic class ID. | `semantic000000.png`, ... |
44
- | **Camera Trajectories** | Flattened 4×4 transformation matrices representing camera poses for each frame. | `traj.txt` (one 4×4 matrix per line) |
45
- | **Question-Answer Pairs** | Validated question-answer pairs related to the scene, optionally associated with specific frames. | `validated_questions.json` |
46
-
47
-
48
- ## VQA Question Categories
49
- The dataset includes a automatically generated answer-question pairs with the following question types:
50
- 1. **Binary General** – Yes/No questions about the presence of objects and general scene characteristics
51
- *Example:* `Is there a blue sofa?`
52
-
53
- 2. **Binary Existence-Based** Yes/No questions designed to track false positives by querying non-existent objects
54
- *Example:* `Is there a piano?`
55
-
56
- 3. **Binary Logical** Yes/No questions with logical operators such as AND/OR
57
- *Example:* `Is there a chair AND a table?`
58
-
59
- 4. **Measurement** Questions requiring numerical answers related to object counts or scene attributes
60
- *Example:* `How many windows are present?`
61
-
62
- 5. **Object Attributes** – Queries about object properties, including color, shape, and material
63
- *Example:* `What color is the door?`
64
-
65
- 6. **Object Relations (Functional)** Questions about functional relationships between objects
66
- *Example:* `Which object supports the table?`
67
-
68
- 7. **Object Relations (Spatial)** – Queries about spatial placement of objects within the scene
69
- *Example:* `What is in front of the staircase?`
70
-
71
- 8. **Comparison** – Questions that compare object properties such as size, color, and position
72
- *Example:* `Which is taller: the bookshelf or the lamp?`
73
-
74
-
75
- ## Citing OSMa-Bench Dataset
76
-
77
- Using OSMa-Bench dataset in your research? Please cite following paper: [OSMa-Bench arxiv](https://arxiv.org/abs/2503.10331).
78
-
79
- ```bibtex
80
- @inproceedings{popov2025osmabench,
81
- title = {OSMa-Bench: Evaluating Open Semantic Mapping Under Varying Lighting Conditions},
82
- author = {Popov, Maxim and Kurkova, Regina and Iumanov, Mikhail and Mahmoud, Jaafar and Kolyubin, Sergey},
83
- booktitle = {2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
84
- year = {2025}
85
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
86
  ```
 
1
+ ---
2
+ license: cc-by-4.0
3
+ ---
4
+
5
+ # OSMa-Bench Dataset
6
+
7
+ [![OSMa-Bench pipeline](https://be2rlab.github.io/OSMa-Bench/static/images/full_pipeline.png)](https://be2rlab.github.io/OSMa-Bench/)
8
+
9
+ OSMa-Bench (Open Semantic Mapping Benchmark) dataset is a fully automatically generated dataset for evaluating the robustness of open semantic mapping and segmentation systems under varying indoor lighting conditions and robot movement dynamics. This dataset is part of [OSMa-Bench](https://be2rlab.github.io/OSMa-Bench/) pipeline.
10
+
11
+ - **Homepage:** [OSMa-Bench Project Page](https://be2rlab.github.io/OSMa-Bench/)
12
+
13
+ ## Dataset Summary
14
+
15
+ This dataset provides simulated RGB-D and semantically annotated posed sequences for evaluation of semantic mapping and segmentation, with a particular focus on handling dynamic lighting—a critical but often overlooked factor in existing benchmarks. It also includes a collection of automatically generated question–answer pairs across multiple categories to support the evaluation of scene-graph–based reasoning, offering a task-driven measure of how well a system’s reconstructed scene captures semantic relationships between objects.
16
+
17
+ The data is built upon two base datasets:
18
+ - **ReplicaCAD**: 22 scenes with 4 lighting configurations and a velocity modifier.
19
+ - **Habitat Matterport 3D (HM3D)**: 8 scenes with 2 lighting configurations and a velocity modifier.
20
+
21
+ ## Installation
22
+
23
+ We offer two versions of the dataset: one with separate files and one as a single compressed archive. Use the following command to download the separated files (may be slow):
24
+
25
+ ```bash
26
+ git xet install
27
+ git clone https://huggingface.co/datasets/warmhammer/OSMa-Bench_dataset -b main
28
+ ```
29
+ and this command to download the compressed version (faster one):
30
+
31
+ ```bash
32
+ git xet install
33
+ git clone https://huggingface.co/datasets/warmhammer/OSMa-Bench_dataset -b compressed
34
+ unzip data.zip
35
+ ```
36
+
37
+ ## Data Configurations
38
+
39
+ The dataset includes the following configurations for the ReplicaCAD and HM3D scenes:
40
+
41
+ | Configuration | Description |
42
+ | :--- | :--- |
43
+ | `baseline` | Static, non-uniformly distributed light sources (ReplicaCAD only) |
44
+ | `dynamic_lighting` | Lighting conditions change along the robot's path (ReplicaCAD only) |
45
+ | `nominal_lights` | The mesh itself emits light without added light sources |
46
+ | `camera_light` | An extra directed light source is attached to the camera |
47
+ | `velocity` | Sequences recorded at doubled nominal velocity |
48
+
49
+ ---
50
+
51
+ ## Data Configurations
52
+
53
+ The dataset provides structured data for each scene, suitable for tasks like 3D scene understanding, visual question answering, and robotics. Each scene contains the following components:
54
+
55
+ | Component | Description | Format / Example |
56
+ | ------------------------- | ------------------------------------------------------------------------------------------------- | --------------------------------------|
57
+ | **RGB Images** | Standard color images captured from different camera viewpoints. | `frame000000.jpg`, ... |
58
+ | **Depth Images** | Depth maps aligned with RGB images. Each pixel encodes depth in meters. | `depth000000.png`, ... |
59
+ | **Semantic Masks** | Pixel-wise semantic segmentation labels. Each pixel corresponds to a semantic class ID. | `semantic000000.png`, ... |
60
+ | **Camera Trajectories** | Flattened 4×4 transformation matrices representing camera poses for each frame. | `traj.txt` (one 4×4 matrix per line) |
61
+ | **Question-Answer Pairs** | Validated question-answer pairs related to the scene, optionally associated with specific frames. | `validated_questions.json` |
62
+
63
+
64
+ ## VQA Question Categories
65
+ The dataset includes a automatically generated answer-question pairs with the following question types:
66
+ 1. **Binary General** Yes/No questions about the presence of objects and general scene characteristics
67
+ *Example:* `Is there a blue sofa?`
68
+
69
+ 2. **Binary Existence-Based** Yes/No questions designed to track false positives by querying non-existent objects
70
+ *Example:* `Is there a piano?`
71
+
72
+ 3. **Binary Logical** Yes/No questions with logical operators such as AND/OR
73
+ *Example:* `Is there a chair AND a table?`
74
+
75
+ 4. **Measurement** Questions requiring numerical answers related to object counts or scene attributes
76
+ *Example:* `How many windows are present?`
77
+
78
+ 5. **Object Attributes** – Queries about object properties, including color, shape, and material
79
+ *Example:* `What color is the door?`
80
+
81
+ 6. **Object Relations (Functional)** Questions about functional relationships between objects
82
+ *Example:* `Which object supports the table?`
83
+
84
+ 7. **Object Relations (Spatial)** – Queries about spatial placement of objects within the scene
85
+ *Example:* `What is in front of the staircase?`
86
+
87
+ 8. **Comparison** – Questions that compare object properties such as size, color, and position
88
+ *Example:* `Which is taller: the bookshelf or the lamp?`
89
+
90
+
91
+ ## Citing OSMa-Bench Dataset
92
+
93
+ Using OSMa-Bench dataset in your research? Please cite following paper: [OSMa-Bench arxiv](https://arxiv.org/abs/2503.10331).
94
+
95
+ ```bibtex
96
+ @inproceedings{popov2025osmabench,
97
+ title = {OSMa-Bench: Evaluating Open Semantic Mapping Under Varying Lighting Conditions},
98
+ author = {Popov, Maxim and Kurkova, Regina and Iumanov, Mikhail and Mahmoud, Jaafar and Kolyubin, Sergey},
99
+ booktitle = {2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
100
+ year = {2025}
101
+ }
102
  ```