J3rr1 commited on
Commit
81fe67e
Β·
verified Β·
1 Parent(s): fd9f7f1

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +49 -0
README.md CHANGED
@@ -2,6 +2,22 @@
2
  license: cc-by-nc-sa-4.0
3
  ---
4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5
 
6
  # Get started πŸ”₯
7
  ## Download the Dataset
@@ -12,9 +28,42 @@ from huggingface_hub import snapshot_download
12
  dataset_path = snapshot_download("InternRobotics/IROS-2025-Challenge-Manip", repo_type="dataset")
13
  ```
14
 
 
 
 
 
 
 
 
 
 
 
15
 
16
  ## Dataset Structure
17
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
18
  ### train Folder hierarchy
19
  ```
20
  train
 
2
  license: cc-by-nc-sa-4.0
3
  ---
4
 
5
+ # IROS-2025-Challenge-Manip
6
+ # Dataset Summary πŸ“–
7
+
8
+ This dataset contains the **IROS Challenge - Manipulation Track** benchmark, organized into **pretrain**, **train**, and **validation** splits.
9
+
10
+ * **Pretrain split**: \~20,000 single pick-and-place trajectories, packaged into tar files (each containing \~1,000 trajectories).
11
+ * **Train split**: task-specific demonstrations, with \~100 trajectories provided per task.
12
+ * **Validation split**: includes the test-time scenes and object assets in **USD format**.
13
+
14
+ Each trajectory in the pretrain and train splits contains:
15
+
16
+ * **Multi-view video** recordings (three perspectives: head-mounted camera and two wrist cameras)
17
+ * **Robot states** (joint positions, gripper states, etc.)
18
+ * **Actions** corresponding to the task execution
19
+
20
+ This dataset is designed to support **pretraining, task-specific fine-tuning, and evaluation** for robotic manipulation in the IROS Challenge setting.
21
 
22
  # Get started πŸ”₯
23
  ## Download the Dataset
 
28
  dataset_path = snapshot_download("InternRobotics/IROS-2025-Challenge-Manip", repo_type="dataset")
29
  ```
30
 
31
+ ## Unzip the pretrain dataset
32
+
33
+ ```bash
34
+ cd pretrain
35
+ for i in {1..20}; do
36
+ echo "Extracting $i.tar.gz ..."
37
+ tar -xzf "$i.tar.gz"
38
+ done
39
+ ```
40
+
41
 
42
  ## Dataset Structure
43
 
44
+ ### pretrain Folder hierarchy
45
+ ```
46
+ pretrain
47
+ β”œβ”€β”€ 1.tar.gz
48
+ β”‚ └── 1/
49
+ β”‚ β”œβ”€β”€ data/
50
+ β”‚ β”œβ”€β”€ meta/
51
+ β”‚ └── videos/
52
+ β”œβ”€β”€ 2.tar.gz
53
+ β”‚ └── 2/
54
+ β”‚ β”œβ”€β”€ data/
55
+ β”‚ β”œβ”€β”€ meta/
56
+ β”‚ └── videos/
57
+ ...
58
+ β”œβ”€β”€ 20.tar.gz
59
+ └── 20/
60
+ β”œβ”€β”€ data/
61
+ β”œβ”€β”€ meta/
62
+ └── videos/
63
+
64
+ ```
65
+
66
+
67
  ### train Folder hierarchy
68
  ```
69
  train