File size: 1,999 Bytes
f595a29
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
---
pretty_name: PAM19
language:
- en
tags:
- robotics
- tactile-sensing
- imitation-learning
- franka-panda
- open-x-tactile
---

# PAM19

PAM19 is a real-world Franka Panda tactile-manipulation dataset containing
1,904 trajectories across 19 tasks. Demonstrations were collected in joint
space with two RGB cameras and two Xense vision-based tactile sensors, then
converted to FTP-1-compatible Zarr format.

The dataset follows the
[Open-X-Tactile](https://open-x-tactile.github.io/) / FTP-1 data convention.

## Dataset summary

- Robot: Franka Panda with a parallel gripper
- Tactile sensor: Xense, two gripper-side image streams
- RGB streams: main camera and wrist camera
- Tasks: 19
- Trajectories: 1,904
- Frames: 2,061,059
- Control/state representation: 7 arm joints and one gripper-opening scalar
- Image resolution in Zarr: 224 × 224

## Tasks

- `USB_plug`
- `cable_routing`
- `cut`
- `cut_long`
- `drop`
- `erase`
- `network_cable`
- `peelCucumber`
- `pick`
- `pick_easy`
- `plug_in`
- `pour`
- `screw_vitmain`
- `screw_xuebi`
- `screw_yibao`
- `screw_yuanqi`
- `tiny_pickA`
- `tiny_pickAB`
- `tiny_pickB`

## Files

Each task directory contains one store-only split ZIP archive of an
FTP-1-compatible `<task>.zarr` directory. The split size is 10 GiB. Small
tasks may contain only a `.zip` file; larger tasks also contain `.z01`,
`.z02`, and subsequent parts.

To reconstruct and extract a split archive:

```bash
zip -s 0 <task>/<task>.zip --out <task>/<task>-full.zip
unzip <task>/<task>-full.zip
```

## Main Zarr fields

- `timestamps`
- `camera_main_rgb`
- `right_wrist_camera_rgb`
- `right_arm_joints`
- `right_hand_joints`
- `right_hand_joints_idx`
- `right_tactile_data_gripper`
- `right_tactile_area_gripper`
- `right_tactile_sensor_gripper`
- `right_tactile_type_gripper`
- `sub_task_instruction`

The gripper scalar uses canonical hand slot 28. The two tactile channels are
stored in left/right order with functional-area indices `[0, 1]`.

## Contact

e1373933@u.nus.edu