djkesu commited on
Commit
307a100
·
verified ·
1 Parent(s): dceef16

Create t-shirt folding manifest dataset

Browse files
Files changed (2) hide show
  1. README.md +81 -0
  2. data/train.jsonl +5 -0
README.md ADDED
@@ -0,0 +1,81 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ pretty_name: T-Shirt Folding Mixed Manifest
3
+ tags:
4
+ - robotics
5
+ - imitation-learning
6
+ - lerobot
7
+ - openpi
8
+ task_categories:
9
+ - other
10
+ size_categories:
11
+ - n<1K
12
+ ---
13
+
14
+ # T-Shirt Folding Mixed Manifest
15
+
16
+ This repo is a lightweight manifest for a public T-shirt folding collection built from five source datasets:
17
+
18
+ - `Gongsta/trlc_tshirt_folding`
19
+ - `Gongsta/trlc_tshirt_folding_impedance`
20
+ - `Gongsta/dagger_dk1_tshirt_corrections`
21
+ - `Gongsta/krish-simpler-tshirt`
22
+ - `Gongsta/e7-tshirt-folding`
23
+
24
+ The goal is to provide one clean public entrypoint while preserving each source dataset at its highest native frame rate.
25
+
26
+ ## What Is In This Repo
27
+
28
+ Each row in `data/train.jsonl` describes one source dataset and includes:
29
+
30
+ - `source_repo`
31
+ - `source_tag`
32
+ - `native_fps`
33
+ - `is_dagger`
34
+ - `is_impedance`
35
+ - `scene_type`
36
+ - `shirt_layout`
37
+ - `recommended_common_fps`
38
+
39
+ This repo does **not** duplicate the source videos. It is a manifest / collection layer that documents provenance and intended loading behavior.
40
+
41
+ ## Source Tags
42
+
43
+ - `apartment_original_multi_shirt`
44
+ - `apartment_impedance_multi_shirt`
45
+ - `multi_shirt_dagger_corrections`
46
+ - `apartment_single_shirt`
47
+ - `building_single_shirt`
48
+
49
+ ## Recommended Loading
50
+
51
+ If you want to combine all five datasets with exact-stride downsampling and no interpolation, use:
52
+
53
+ - `target_fps = 10`
54
+
55
+ Why:
56
+
57
+ - `30 -> 10` is exact stride `3`
58
+ - `50 -> 10` is exact stride `5`
59
+
60
+ If you want to keep native FPS, load the listed source repos directly and treat this dataset as the metadata / provenance index.
61
+
62
+ ## Notes
63
+
64
+ - `Gongsta/trlc_tshirt_folding` is the only `30 Hz` source.
65
+ - `Gongsta/dagger_dk1_tshirt_corrections` is the DAgger dataset and is `50 Hz`.
66
+ - The other listed sources are `50 Hz`.
67
+
68
+ ## Example
69
+
70
+ ```python
71
+ repo_ids = [
72
+ "Gongsta/trlc_tshirt_folding",
73
+ "Gongsta/trlc_tshirt_folding_impedance",
74
+ "Gongsta/dagger_dk1_tshirt_corrections",
75
+ "Gongsta/krish-simpler-tshirt",
76
+ "Gongsta/e7-tshirt-folding",
77
+ ]
78
+
79
+ # For an exact-stride common timebase across all five:
80
+ target_fps = 10
81
+ ```
data/train.jsonl ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ {"source_repo":"Gongsta/trlc_tshirt_folding","source_tag":"apartment_original_multi_shirt","native_fps":30,"is_dagger":false,"is_impedance":false,"scene_type":"apartment","shirt_layout":"multi_shirt","recommended_common_fps":10}
2
+ {"source_repo":"Gongsta/trlc_tshirt_folding_impedance","source_tag":"apartment_impedance_multi_shirt","native_fps":50,"is_dagger":false,"is_impedance":true,"scene_type":"apartment","shirt_layout":"multi_shirt","recommended_common_fps":10}
3
+ {"source_repo":"Gongsta/dagger_dk1_tshirt_corrections","source_tag":"multi_shirt_dagger_corrections","native_fps":50,"is_dagger":true,"is_impedance":false,"scene_type":"mixed","shirt_layout":"multi_shirt","recommended_common_fps":10}
4
+ {"source_repo":"Gongsta/krish-simpler-tshirt","source_tag":"apartment_single_shirt","native_fps":50,"is_dagger":false,"is_impedance":false,"scene_type":"apartment","shirt_layout":"single_shirt","recommended_common_fps":10}
5
+ {"source_repo":"Gongsta/e7-tshirt-folding","source_tag":"building_single_shirt","native_fps":50,"is_dagger":false,"is_impedance":false,"scene_type":"building","shirt_layout":"single_shirt","recommended_common_fps":10}