Pranay Thangeda
commited on
Commit
•
7c432f1
1
Parent(s):
bd33da7
Add scooping dataset
Browse filesThis view is limited to 50 files because it contains too many changes.
See raw diff
- .gitattributes +1 -0
- README.md +147 -0
- depth_images/terrain_10_sample_100_depth.png +3 -0
- depth_images/terrain_10_sample_10_depth.png +3 -0
- depth_images/terrain_10_sample_11_depth.png +3 -0
- depth_images/terrain_10_sample_12_depth.png +3 -0
- depth_images/terrain_10_sample_13_depth.png +3 -0
- depth_images/terrain_10_sample_14_depth.png +3 -0
- depth_images/terrain_10_sample_15_depth.png +3 -0
- depth_images/terrain_10_sample_16_depth.png +3 -0
- depth_images/terrain_10_sample_17_depth.png +3 -0
- depth_images/terrain_10_sample_18_depth.png +3 -0
- depth_images/terrain_10_sample_19_depth.png +3 -0
- depth_images/terrain_10_sample_1_depth.png +3 -0
- depth_images/terrain_10_sample_20_depth.png +3 -0
- depth_images/terrain_10_sample_21_depth.png +3 -0
- depth_images/terrain_10_sample_22_depth.png +3 -0
- depth_images/terrain_10_sample_23_depth.png +3 -0
- depth_images/terrain_10_sample_24_depth.png +3 -0
- depth_images/terrain_10_sample_25_depth.png +3 -0
- depth_images/terrain_10_sample_26_depth.png +3 -0
- depth_images/terrain_10_sample_27_depth.png +3 -0
- depth_images/terrain_10_sample_28_depth.png +3 -0
- depth_images/terrain_10_sample_29_depth.png +3 -0
- depth_images/terrain_10_sample_2_depth.png +3 -0
- depth_images/terrain_10_sample_30_depth.png +3 -0
- depth_images/terrain_10_sample_31_depth.png +3 -0
- depth_images/terrain_10_sample_32_depth.png +3 -0
- depth_images/terrain_10_sample_33_depth.png +3 -0
- depth_images/terrain_10_sample_34_depth.png +3 -0
- depth_images/terrain_10_sample_35_depth.png +3 -0
- depth_images/terrain_10_sample_36_depth.png +3 -0
- depth_images/terrain_10_sample_37_depth.png +3 -0
- depth_images/terrain_10_sample_38_depth.png +3 -0
- depth_images/terrain_10_sample_39_depth.png +3 -0
- depth_images/terrain_10_sample_3_depth.png +3 -0
- depth_images/terrain_10_sample_40_depth.png +3 -0
- depth_images/terrain_10_sample_41_depth.png +3 -0
- depth_images/terrain_10_sample_42_depth.png +3 -0
- depth_images/terrain_10_sample_43_depth.png +3 -0
- depth_images/terrain_10_sample_44_depth.png +3 -0
- depth_images/terrain_10_sample_45_depth.png +3 -0
- depth_images/terrain_10_sample_46_depth.png +3 -0
- depth_images/terrain_10_sample_47_depth.png +3 -0
- depth_images/terrain_10_sample_48_depth.png +3 -0
- depth_images/terrain_10_sample_49_depth.png +3 -0
- depth_images/terrain_10_sample_4_depth.png +3 -0
- depth_images/terrain_10_sample_50_depth.png +3 -0
- depth_images/terrain_10_sample_51_depth.png +3 -0
- depth_images/terrain_10_sample_52_depth.png +3 -0
.gitattributes
CHANGED
@@ -56,3 +56,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
56 |
# Video files - compressed
|
57 |
*.mp4 filter=lfs diff=lfs merge=lfs -text
|
58 |
*.webm filter=lfs diff=lfs merge=lfs -text
|
|
|
|
56 |
# Video files - compressed
|
57 |
*.mp4 filter=lfs diff=lfs merge=lfs -text
|
58 |
*.webm filter=lfs diff=lfs merge=lfs -text
|
59 |
+
*.csv filter=lfs diff=lfs merge=lfs -text
|
README.md
ADDED
@@ -0,0 +1,147 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Scooping Dataset
|
2 |
+
|
3 |
+
This dataset contains 6,700 samples collected over 67 terrains for the task of manipulating granular materials using a robotic arm. The dataset is designed to facilitate research in robotic manipulation, machine learning, and related fields. Each sample includes:
|
4 |
+
|
5 |
+
- **Terrain Metadata**: Terrain ID, composition, and materials used.
|
6 |
+
- **RGB Image**: An RGB image of the terrain before scooping.
|
7 |
+
- **Depth Image**: Depth data corresponding to the terrain saved as a 16-bit PNG image.
|
8 |
+
- **Depth Normalization Parameters**: `depth_min` and `depth_max` to reconstruct original depth values.
|
9 |
+
- **F/T Sensor Data**: Force/Torque sensor data captured during the scooping action, saved as a CSV file.
|
10 |
+
- **Action Parameters**: Scoop location, yaw angle, scoop depth, and stiffness.
|
11 |
+
- **Outcome**: Volume of material scooped.
|
12 |
+
|
13 |
+
## Dataset Structure
|
14 |
+
|
15 |
+
### Features
|
16 |
+
|
17 |
+
- `terrain_id` (`int32`): Terrain identifier (1-67).
|
18 |
+
- `composition` (`string`): Terrain composition (`single`, `partition`, `mixture`, `layers`).
|
19 |
+
- `material_1` (`string`): First material used in the terrain.
|
20 |
+
- `material_2` (`string`): Second material used in the terrain (empty string if not applicable).
|
21 |
+
- `material_3` (`string`): Third material used in the terrain (empty string if not applicable).
|
22 |
+
- `sample_index` (`int32`): Sample index within the terrain (1-100).
|
23 |
+
- `rgb_image` (`Image`): RGB image of the terrain before scooping.
|
24 |
+
- `depth_image` (`Image`): Depth data as a 16-bit PNG image.
|
25 |
+
- `depth_min` (`float32`): Minimum depth value used for normalization.
|
26 |
+
- `depth_max` (`float32`): Maximum depth value used for normalization.
|
27 |
+
- `ft_csv_path` (`string`): File path to the F/T sensor data (CSV file).
|
28 |
+
- `pixel_x` (`float32`): X-coordinate of the scoop location in the image.
|
29 |
+
- `pixel_y` (`float32`): Y-coordinate of the scoop location in the image.
|
30 |
+
- `yaw` (`float32`): Yaw angle of the end-effector (in radians).
|
31 |
+
- `scoop_depth` (`float32`): Depth of the scoop (in meters).
|
32 |
+
- `stiffness` (`float32`): Stiffness of the controller during the scoop action.
|
33 |
+
- `scooped_volume` (`float32`): Volume of material scooped (in cubic meters).
|
34 |
+
|
35 |
+
### Data Files
|
36 |
+
|
37 |
+
The dataset is organized with the following structure:
|
38 |
+
|
39 |
+
```
|
40 |
+
scooping_dataset/
|
41 |
+
├── scooping_dataset.arrow
|
42 |
+
├── dataset_info.json
|
43 |
+
├── rgb_images/
|
44 |
+
│ ├── terrain_1_sample_1_rgb.png
|
45 |
+
│ ├── terrain_1_sample_2_rgb.png
|
46 |
+
│ └── ...
|
47 |
+
├── depth_images/
|
48 |
+
│ ├── terrain_1_sample_1_depth.png
|
49 |
+
│ ├── terrain_1_sample_2_depth.png
|
50 |
+
│ └── ...
|
51 |
+
├── ft_data/
|
52 |
+
│ ├── terrain_1_sample_1_ft.csv
|
53 |
+
│ ├── terrain_1_sample_2_ft.csv
|
54 |
+
│ └── ...
|
55 |
+
```
|
56 |
+
|
57 |
+
## Usage
|
58 |
+
|
59 |
+
To use this dataset, you can load it using the Hugging Face `datasets` library:
|
60 |
+
|
61 |
+
```python
|
62 |
+
from datasets import load_from_disk
|
63 |
+
import numpy as np
|
64 |
+
from PIL import Image as PILImage
|
65 |
+
import pandas as pd
|
66 |
+
|
67 |
+
# Load the dataset
|
68 |
+
dataset = load_from_disk('path_to_dataset/scooping_dataset')
|
69 |
+
|
70 |
+
# Access a sample
|
71 |
+
sample = dataset[0]
|
72 |
+
|
73 |
+
# Load RGB image
|
74 |
+
rgb_image = sample['rgb_image'] # PIL Image
|
75 |
+
rgb_image.show()
|
76 |
+
|
77 |
+
# Load depth image and reconstruct original depth values
|
78 |
+
depth_image = sample['depth_image'] # PIL Image
|
79 |
+
depth_array = np.array(depth_image).astype(np.float32)
|
80 |
+
depth_normalized = depth_array / 65535 # Normalize back to [0, 1]
|
81 |
+
depth_min = sample['depth_min']
|
82 |
+
depth_max = sample['depth_max']
|
83 |
+
original_depth = depth_normalized * (depth_max - depth_min) + depth_min
|
84 |
+
|
85 |
+
# Load F/T sensor data
|
86 |
+
ft_data = pd.read_csv(sample['ft_csv_path'], header=None).values
|
87 |
+
|
88 |
+
# Access action parameters
|
89 |
+
pixel_x = sample['pixel_x']
|
90 |
+
pixel_y = sample['pixel_y']
|
91 |
+
yaw = sample['yaw']
|
92 |
+
scoop_depth = sample['scoop_depth']
|
93 |
+
stiffness = sample['stiffness']
|
94 |
+
scooped_volume = sample['scooped_volume']
|
95 |
+
```
|
96 |
+
|
97 |
+
## Dataset Creation
|
98 |
+
|
99 |
+
The dataset was created using the following steps:
|
100 |
+
|
101 |
+
1. **Data Collection**: 6,700 samples were collected using a UR5e robotic arm equipped with a scooping end-effector and an overhead RealSense L515 camera to capture RGB-D images.
|
102 |
+
|
103 |
+
2. **Terrain Preparation**: 67 terrains were prepared using combinations of 12 different materials and 4 types of compositions. Each terrain represents a unique task.
|
104 |
+
|
105 |
+
3. **Action Execution**: For each terrain, 100 scooping actions were performed with varying action parameters (scoop location, yaw, depth, and stiffness).
|
106 |
+
|
107 |
+
4. **Data Recording**: Before each action, an RGB-D image of the terrain was captured. During the action, F/T sensor data was recorded. After the action, the volume of material scooped was measured.
|
108 |
+
|
109 |
+
## License
|
110 |
+
|
111 |
+
This dataset is licensed under the **Creative Commons Attribution 4.0 International (CC BY 4.0) License**.
|
112 |
+
|
113 |
+
[![License: CC BY 4.0](https://licensebuttons.net/l/by/4.0/88x31.png)](https://creativecommons.org/licenses/by/4.0/)
|
114 |
+
|
115 |
+
You are free to:
|
116 |
+
|
117 |
+
- **Share** — copy and redistribute the material in any medium or format.
|
118 |
+
- **Adapt** — remix, transform, and build upon the material for any purpose, even commercially.
|
119 |
+
|
120 |
+
Under the following terms:
|
121 |
+
|
122 |
+
- **Attribution** — You must give appropriate credit, provide a link to the license, and indicate if changes were made.
|
123 |
+
|
124 |
+
**Full License Text**: [https://creativecommons.org/licenses/by/4.0/legalcode](https://creativecommons.org/licenses/by/4.0/legalcode)
|
125 |
+
|
126 |
+
## Citation
|
127 |
+
|
128 |
+
If you use this dataset, please cite the following paper:
|
129 |
+
|
130 |
+
```bibtex
|
131 |
+
@inproceedings{Zhu-RSS-23,
|
132 |
+
author = {Zhu, Yifan and Thangeda, Pranay and Ornik, Melkior and Hauser, Kris},
|
133 |
+
title = {Few-shot Adaptation for Manipulating Granular Materials Under Domain Shift},
|
134 |
+
booktitle = {Proceedings of Robotics: Science and Systems},
|
135 |
+
year = {2023},
|
136 |
+
month = {July},
|
137 |
+
address = {Daegu, Republic of Korea},
|
138 |
+
doi = {10.15607/RSS.2023.XIX.048}
|
139 |
+
}
|
140 |
+
```
|
141 |
+
|
142 |
+
## References
|
143 |
+
|
144 |
+
For more details and illustrations of the materials and compositions, please visit our [project website](https://drillaway.github.io/).
|
145 |
+
|
146 |
+
---
|
147 |
+
|
depth_images/terrain_10_sample_100_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_10_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_11_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_12_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_13_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_14_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_15_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_16_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_17_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_18_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_19_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_1_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_20_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_21_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_22_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_23_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_24_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_25_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_26_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_27_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_28_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_29_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_2_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_30_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_31_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_32_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_33_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_34_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_35_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_36_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_37_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_38_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_39_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_3_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_40_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_41_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_42_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_43_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_44_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_45_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_46_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_47_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_48_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_49_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_4_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_50_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_51_depth.png
ADDED
Git LFS Details
|
depth_images/terrain_10_sample_52_depth.png
ADDED
Git LFS Details
|