Datasets:

Modalities:
Image
Size:
< 1K
ArXiv:
Libraries:
Datasets
License:
tillspeicher commited on
Commit
57aafdc
·
verified ·
1 Parent(s): 3bc04bc

Upload folder using huggingface_hub

Browse files
Files changed (2) hide show
  1. README.md +42 -10
  2. samples_attribution.md +11 -0
README.md CHANGED
@@ -8,21 +8,53 @@ configs:
8
  ---
9
  # Transforms-2D Base Dataset
10
 
11
- Foreground objects and backgrounds used by the Transforms-2D dataset.
 
12
 
13
- ## Attribution
 
14
 
15
- The data here is based on the [SI-Score dataset](https://github.com/google-research/si-score/tree/master?tab=readme-ov-file) ([paper](https://arxiv.org/abs/2007.08558)) and re-uploaded to HF to make it easier to access than the original AWS S3 bucket.
16
- If you are using this dataset, please consider citing the original authors as well:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
17
 
 
 
 
18
  ```bibtex
19
- @misc{djolonga2020robustness,
20
- title={On Robustness and Transferability of Convolutional Neural Networks},
21
- author={Josip Djolonga and Jessica Yung and Michael Tschannen and Rob Romijnders and Lucas Beyer and Alexander Kolesnikov and Joan Puigcerver and Matthias Minderer and Alexander D'Amour and Dan Moldovan and Sylvain Gelly and Neil Houlsby and Xiaohua Zhai and Mario Lucic},
22
- year={2020},
23
- eprint={2007.08558},
24
  archivePrefix={arXiv},
25
- primaryClass={cs.CV}
 
26
  }
27
  ```
28
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8
  ---
9
  # Transforms-2D Base Dataset
10
 
11
+ This dataset contains foreground objects and background images used by the Transforms-2D dataset in the paper [Understanding the Role of Invariance in Transfer Learning](https://arxiv.org/abs/2407.04325), published at TMLR 2024.
12
+ The code for the paper is available [here](https://github.com/tillspeicher/representation-invariance-transfer), including the [implementation of the Transforms-2D dataset](https://github.com/tillspeicher/representation-invariance-transfer/tree/master/src/transforms_2d).
13
 
14
+ The Transforms-2D dataset consists of transformed versions of image objects with transparency masks (from this base dataset), pasted onto background images (also from this base dataset).
15
+ It is used to study the role of invariance in transfer learning, by creating images with carefully controlled transformations.
16
 
17
+ ## Usage
18
+
19
+ The dataset here comes in two configurations: a `foregrounds` configuration with 61 classes of images and several images per class, and a `backgrounds` configuration with 867 background images of nature scenes.
20
+
21
+ To load the respective configuration, use
22
+
23
+ ```python
24
+ from datasets import load_dataset
25
+
26
+ data = load_dataset(
27
+ "tillspeicher/transforms_2d_base",
28
+ "foregrounds", # or "backgrounds"
29
+ # There's only one the "train" split
30
+ split="train",
31
+ )
32
+ ```
33
 
34
+ ## Citation
35
+
36
+ If you are using the Transform-2D dataset, please consider citing the following paper:
37
  ```bibtex
38
+ @misc{speicher2024understanding,
39
+ title={Understanding the Role of Invariance in Transfer Learning},
40
+ author={Till Speicher and Vedant Nanda and Krishna P. Gummadi},
41
+ year={2024},
42
+ eprint={2407.04325},
43
  archivePrefix={arXiv},
44
+ primaryClass={cs.LG},
45
+ url={https://arxiv.org/abs/2407.04325},
46
  }
47
  ```
48
 
49
+
50
+ ## Attribution
51
+
52
+ The data here is based on the [SI-Score dataset](https://github.com/google-research/si-score/tree/master?tab=readme-ov-file) ([paper](https://arxiv.org/abs/2007.08558)) and re-uploaded to HF to make it easier to access than the original AWS S3 bucket.
53
+ If you are using this dataset, please consider citing the original authors as well.
54
+
55
+ The foreground images are segmented versions of OpenImages, with CC-licenses.
56
+ The attributions for each image can be found on the [OpenImages](https://storage.googleapis.com/openimages/web/download.html) website in the Image IDs CSVs.
57
+
58
+ The background images come from Pexels.com and carry a Pexels license.
59
+ Some of the background images do not carry a Pexels license.
60
+ The attributions for these images are listed in `samples_attributions.md`.
samples_attribution.md ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Here are the attributions for the background images in `background_samples/`.
2
+
3
+ - beach2: [Beach in Pagudpud, Ilocos Norte, Philippines]() by John Ryan Cordova from Philippines. (Creative Commons Attribution-Share Alike 2.0 Generic license)
4
+ - field1: [Pelto Kärkölässä](https://sq.m.wikipedia.org/wiki/Skeda:Field_in_K%C3%A4rk%C3%B6l%C3%A4.jpg) by Okko Pyykkö
5
+ (Creative Commons Attribution 2.0 Generic license)
6
+ - forest1: [Photo](https://pxhere.com/en/photo/945157) on Pxhere (CC0 license).
7
+ - grass1: [A LOT OF GREEN GRASS ON THE GROUND](https://pixy.org/4783081/) by Anne Fonda (CC0 Public Domain license).
8
+ - ocean1: [Dead calm at sea](https://www.flickr.com/photos/gaelvaroquaux/34212317084) by Gael Varoquaux (CC BY 2.0 Generic license).
9
+ - ocean3: [okyanus-deniz-sonsuz-doğa-yaz](https://pixabay.com/tr/photos/okyanus-deniz-sonsuz-do%C4%9Fa-yaz-3264052/) by TeeFarm (Pixabay license).
10
+ - mountains_blue_sky: [Pink Flowers Near Mountain Covered by Snow](https://www.pexels.com/photo/landscape-nature-night-relaxation-36478/) by Pixabay (CC0 license).
11
+ - sky1: [Blue Skies](https://www.pexels.com/photo/nature-sky-clouds-blue-53594/) by Pixabay (CC0 license).