Datasets:

Modalities:
Image
Formats:
parquet
ArXiv:
Libraries:
Datasets
pandas
License:
adamnarozniak commited on
Commit
44c23ba
1 Parent(s): 8e3c2c3

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +158 -42
README.md CHANGED
@@ -1,42 +1,158 @@
1
- ---
2
- license: cc-by-4.0
3
- dataset_info:
4
- features:
5
- - name: image
6
- dtype: image
7
- - name: label
8
- dtype:
9
- class_label:
10
- names:
11
- '0': airplane
12
- '1': automobile
13
- '2': bird
14
- '3': cat
15
- '4': deer
16
- '5': dog
17
- '6': frog
18
- '7': horse
19
- '8': ship
20
- '9': truck
21
- splits:
22
- - name: train
23
- num_bytes: 178662714.0
24
- num_examples: 90000
25
- - name: validation
26
- num_bytes: 180126542.0
27
- num_examples: 90000
28
- - name: test
29
- num_bytes: 178913694.0
30
- num_examples: 90000
31
- download_size: 771149160
32
- dataset_size: 537702950.0
33
- configs:
34
- - config_name: default
35
- data_files:
36
- - split: train
37
- path: data/train-*
38
- - split: validation
39
- path: data/validation-*
40
- - split: test
41
- path: data/test-*
42
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-4.0
3
+ dataset_info:
4
+ features:
5
+ - name: image
6
+ dtype: image
7
+ - name: label
8
+ dtype:
9
+ class_label:
10
+ names:
11
+ '0': airplane
12
+ '1': automobile
13
+ '2': bird
14
+ '3': cat
15
+ '4': deer
16
+ '5': dog
17
+ '6': frog
18
+ '7': horse
19
+ '8': ship
20
+ '9': truck
21
+ splits:
22
+ - name: train
23
+ num_bytes: 178662714
24
+ num_examples: 90000
25
+ - name: validation
26
+ num_bytes: 180126542
27
+ num_examples: 90000
28
+ - name: test
29
+ num_bytes: 178913694
30
+ num_examples: 90000
31
+ download_size: 771149160
32
+ dataset_size: 537702950
33
+ configs:
34
+ - config_name: default
35
+ data_files:
36
+ - split: train
37
+ path: data/train-*
38
+ - split: validation
39
+ path: data/validation-*
40
+ - split: test
41
+ path: data/test-*
42
+ task_categories:
43
+ - image-classification
44
+ size_categories:
45
+ - 100K<n<1M
46
+ ---
47
+ # Dataset Card for CINIC-10
48
+
49
+ CINIC-10 has a total of 270,000 images equally split amongst three subsets: train, validate, and test. This means that CINIC-10 has 4.5 times as many samples than CIFAR-10.
50
+
51
+ ## Dataset Details
52
+
53
+ In each subset (90,000 images), there are ten classes (identical to [CIFAR-10](https://www.cs.toronto.edu/~kriz/cifar.html) classes). There are 9000 images per class per subset. Using the suggested data split (an equal three-way split), CINIC-10 has 1.8 times as many training samples as in CIFAR-10. CINIC-10 is designed to be directly swappable with CIFAR-10.
54
+ To understand the motivation behind the dataset creation please visit the [GitHub repository](https://github.com/BayesWatch/cinic-10 ).
55
+
56
+ ### Dataset Sources
57
+
58
+ - **Repository:** https://github.com/BayesWatch/cinic-10
59
+ - **Paper:** https://arxiv.org/abs/1810.03505
60
+ - **Dataset:** http://dx.doi.org/10.7488/ds/2448
61
+ - **Benchmarking, Papers with code:** https://paperswithcode.com/sota/image-classification-on-cinic-10
62
+
63
+ ## Use in FL
64
+
65
+ In order to prepare the dataset for the FL settings, we recommend using [Flower Dataset](https://flower.ai/docs/datasets/) (flwr-datasets) for the dataset download and partitioning and [Flower](https://flower.ai/docs/framework/) (flwr) for conducting FL experiments.
66
+
67
+ To partition the dataset, do the following.
68
+ 1. Install the package.
69
+ ```bash
70
+ pip install flwr-datasets[vision]
71
+ ```
72
+ 2. Use the HF Dataset under the hood in Flower Datasets.
73
+ ```python
74
+ from flwr_datasets import FederatedDataset
75
+ from flwr_datasets.partitioner import IidPartitioner
76
+
77
+ fds = FederatedDataset(
78
+ dataset="flwrlabs/cinic10",
79
+ partitioners={"train": IidPartitioner(num_partitions=10)}
80
+ )
81
+ partition = fds.load_partition(partition_id=0)
82
+ ```
83
+
84
+ ## Dataset Structure
85
+
86
+ ### Data Instances
87
+ The first instance of the train split is presented below:
88
+ ```
89
+ {
90
+ 'image': <PIL.PngImagePlugin.PngImageFile image mode=RGB size=32x32>,
91
+ 'label': 0
92
+ }
93
+ ```
94
+ ### Data Split
95
+
96
+ ```
97
+ DatasetDict({
98
+ train: Dataset({
99
+ features: ['image', 'label'],
100
+ num_rows: 90000
101
+ })
102
+ validation: Dataset({
103
+ features: ['image', 'label'],
104
+ num_rows: 90000
105
+ })
106
+ test: Dataset({
107
+ features: ['image', 'label'],
108
+ num_rows: 90000
109
+ })
110
+ })
111
+ ```
112
+
113
+ ## Citation
114
+
115
+ When working with the CINIC-10 dataset, please cite the original paper.
116
+ If you're using this dataset with Flower Datasets and Flower, cite Flower.
117
+
118
+ **BibTeX:**
119
+
120
+ Original paper:
121
+ ```
122
+ @misc{darlow2018cinic10imagenetcifar10,
123
+ title={CINIC-10 is not ImageNet or CIFAR-10},
124
+ author={Luke N. Darlow and Elliot J. Crowley and Antreas Antoniou and Amos J. Storkey},
125
+ year={2018},
126
+ eprint={1810.03505},
127
+ archivePrefix={arXiv},
128
+ primaryClass={cs.CV},
129
+ url={https://arxiv.org/abs/1810.03505},
130
+ }
131
+ ````
132
+
133
+ Flower:
134
+
135
+ ```
136
+ @article{DBLP:journals/corr/abs-2007-14390,
137
+ author = {Daniel J. Beutel and
138
+ Taner Topal and
139
+ Akhil Mathur and
140
+ Xinchi Qiu and
141
+ Titouan Parcollet and
142
+ Nicholas D. Lane},
143
+ title = {Flower: {A} Friendly Federated Learning Research Framework},
144
+ journal = {CoRR},
145
+ volume = {abs/2007.14390},
146
+ year = {2020},
147
+ url = {https://arxiv.org/abs/2007.14390},
148
+ eprinttype = {arXiv},
149
+ eprint = {2007.14390},
150
+ timestamp = {Mon, 03 Aug 2020 14:32:13 +0200},
151
+ biburl = {https://dblp.org/rec/journals/corr/abs-2007-14390.bib},
152
+ bibsource = {dblp computer science bibliography, https://dblp.org}
153
+ }
154
+ ```
155
+
156
+ ## Dataset Card Contact
157
+
158
+ If you have any questions about the dataset preprocessing and preparation, please contact [Flower Labs](https://flower.ai/).