Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
task_categories:
|
3 |
+
- robotics
|
4 |
+
tags:
|
5 |
+
- code
|
6 |
+
size_categories:
|
7 |
+
- 100B<n<1T
|
8 |
+
---
|
9 |
+
# Robotic Manipulation Datasets for Four Tasks
|
10 |
+
|
11 |
+
[[Project Page]](https://data-scaling-laws.github.io/)
|
12 |
+
[[Paper]](https://data-scaling-laws.github.io/paper.pdf)
|
13 |
+
[[Code]](https://github.com/Fanqi-Lin/Data-Scaling-Laws)
|
14 |
+
[[Models]](https://huggingface.co/Fanqi-Lin/Task-Models/)
|
15 |
+
|
16 |
+
This repository contains in-the-wild robotic manipulation datasets collected using [UMI](https://umi-gripper.github.io/), and processed through a SLAM pipeline, as described in the paper "Data Scaling Laws in Imitation Learning for Robotic Manipulation". The datasets cover four tasks:
|
17 |
+
+ Pour Water
|
18 |
+
+ Arrange Mouse
|
19 |
+
+ Fold Towel
|
20 |
+
+ Unplug Charger
|
21 |
+
|
22 |
+
## Dataset Folders:
|
23 |
+
**arrange_mouse** and **pour_water**: Each folder contains data from 32 unique environment-object pairs, with 120 demonstrations per pair.
|
24 |
+
|
25 |
+
**fold_towel** and **unplug_charger**: Each folder contains data from 32 unique environment-object pairs, with 60 demonstrations per pair.
|
26 |
+
|
27 |
+
**pour_water_16_env_4_object** and **arrange_mouse_16_env_4_object**: These folders contain data from 16 environments, with 4 different manipulation objects per environment, and 120 demonstrations per object.
|
28 |
+
|
29 |
+
These datasets can be used to train policies that generalize effectively to novel environments and objects.
|
30 |
+
|
31 |
+
For more details on how to use our datasets, please refer to our [code](https://github.com/Fanqi-Lin/Data-Scaling-Laws).
|