Datasets:

Modalities:
Text
Video
ArXiv:
License:

Add dataset card, paper link, and sample usage

#1
by nielsr HF Staff - opened
Files changed (2) hide show
  1. .gitattributes +0 -2
  2. README.md +0 -12
.gitattributes CHANGED
@@ -58,5 +58,3 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
58
  # Video files - compressed
59
  *.mp4 filter=lfs diff=lfs merge=lfs -text
60
  *.webm filter=lfs diff=lfs merge=lfs -text
61
- *.ply filter=lfs diff=lfs merge=lfs -text
62
- *.splat filter=lfs diff=lfs merge=lfs -text
 
58
  # Video files - compressed
59
  *.mp4 filter=lfs diff=lfs merge=lfs -text
60
  *.webm filter=lfs diff=lfs merge=lfs -text
 
 
README.md CHANGED
@@ -10,18 +10,6 @@ task_categories:
10
 
11
  **Deform360** is a massive multi-view visuotactile dataset for deformable-object research, featuring 198 daily-life objects, 1,980 interaction sequences, and over 215 hours of observations from 41 surround-view cameras and bimanual tactile grippers to capture both global motion and contact-induced local deformations.
12
 
13
- ## Where the data lives
14
-
15
- | Repository | Contents |
16
- | --- | --- |
17
- | [`brownu/deform360`](https://huggingface.co/datasets/brownu/deform360) (this repo) | `raw/` — the untouched capture directories, one per object |
18
- | [`brownu/deform360_processed`](https://huggingface.co/datasets/brownu/deform360_processed) | `processed/` — undistorted cameras, refined masks, tracking, Gaussian splats and point clouds |
19
-
20
- The processed stage is published in its own repository, in a packed per-episode
21
- layout rather than as loose files. Kept unpacked it runs to well over a million
22
- files, far past the Hub's [per-repository limits](https://huggingface.co/docs/hub/storage-limits)
23
- (under 100k files per repo), which degrades every commit and download against it.
24
-
25
  ## Installation
26
 
27
  To load and interact with the dataset, install the helper library:
 
10
 
11
  **Deform360** is a massive multi-view visuotactile dataset for deformable-object research, featuring 198 daily-life objects, 1,980 interaction sequences, and over 215 hours of observations from 41 surround-view cameras and bimanual tactile grippers to capture both global motion and contact-induced local deformations.
12
 
 
 
 
 
 
 
 
 
 
 
 
 
13
  ## Installation
14
 
15
  To load and interact with the dataset, install the helper library: