Datasets:

Modalities:
Image
Formats:
parquet
ArXiv:
Libraries:
Datasets
Dask
License:
rmokady commited on
Commit
84a37be
·
verified ·
1 Parent(s): 07b9a2f

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +28 -0
README.md CHANGED
@@ -1,4 +1,5 @@
1
  ---
 
2
  dataset_info:
3
  features:
4
  - name: image
@@ -15,3 +16,30 @@ configs:
15
  - split: train
16
  path: data/train-*
17
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ license: apache-2.0
3
  dataset_info:
4
  features:
5
  - name: image
 
16
  - split: train
17
  path: data/train-*
18
  ---
19
+
20
+
21
+
22
+ ## This repository contains Unofficial access for SDIP-dogs dataset
23
+
24
+
25
+ [Official Repository](https://github.com/self-distilled-stylegan/self-distilled-internet-photos), [Project Page](https://self-distilled-stylegan.github.io/), [Paper](https://arxiv.org/abs/2202.12211)
26
+
27
+
28
+ Self-Distilled Internet Photos (SDIP) is a multi-domain image dataset. The dataset consists of Self-Distilled Flickr (SD-Flickr) and *Self-Distilled LSUN (SD-LSUN) that were crawled from [Flickr](https://www.flickr.com/) and [LSUN dataset](https://www.yf.io/p/lsun), respectively, and then curated using the method described in our Self-Distilled StyleGAN paper:
29
+
30
+ > **Self-Distilled StyleGAN: Towards Generation from Internet Photos**<br>
31
+ > Ron Mokady, Michal Yarom, Omer Tov, Oran Lang, Daniel Cohen-Or, Tali Dekel, Michal Irani, Inbar Mosseri
32
+
33
+
34
+ ![](SDIP.png)
35
+
36
+ ## Overview
37
+
38
+ [StyleGAN’s](https://github.com/NVlabs/stylegan2-ada-pytorch) fascinating generative and editing abilities are limited to structurally aligned and well-curated datasets. It does not work well on raw datasets downloaded from the Internet. The SDIP domains presented here, which are StyleGAN-friendly, were automatically curated by our [method](https://arxiv.org/abs/2202.12211) from raw images collected from the Internet. The raw uncurated images in *Self-Distilled Flicker (SD-Flickr)* were first crawled from [Flickr](https://www.flickr.com/) using a simple keyword (e.g. 'dog' or 'elephant').
39
+
40
+ The dataset in this page exhibits 4 domains: **SD-Dogs (126K images) 1024X1024**, **SD-Elephants (39K images) 512X512**, **SD-Bicycles (96K images) 256X256**, and **SD-Horses (162K images) 256X256**. Our curation process consists of a simple pre-processing step (off-the-shelf object detector to crop the main object and then rescale), followed by a sophisticated StyleGAN-friendly filtering step (which removes outlier images while maintaining dataset diversity). This results in a more coherent and clean dataset, which is suitable for training a StyleGAN2 generator (see more details in our [paper](https://arxiv.org/abs/2202.12211)).
41
+
42
+ The data itself is saved in a json format: for SD-Flickr we provide urls of the original images and bounding boxes used for cropping; for SD-LSUN we provide image identifiers with the bounding boxes. In addition to the SDIP dataset, we also provide weights of pre-trained StyleGAN2 models trained using each image domain presented in the paper.
43
+
44
+
45
+