Datasets:

Modalities:
Image
Formats:
parquet
ArXiv:
Libraries:
Datasets
pandas
License:
adamnarozniak commited on
Commit
a391e37
1 Parent(s): 34cfc38

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +112 -0
README.md CHANGED
@@ -33,4 +33,116 @@ configs:
33
  path: data/train-*
34
  - split: test
35
  path: data/test-*
 
 
 
 
 
36
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
33
  path: data/train-*
34
  - split: test
35
  path: data/test-*
36
+ license: unknown
37
+ task_categories:
38
+ - image-classification
39
+ size_categories:
40
+ - 1K<n<10K
41
  ---
42
+
43
+ # Dataset Card for USPS
44
+
45
+ USPS is a digit dataset automatically scanned from envelopes by the U.S. Postal Service containing a total of 9,298 16×16 pixel grayscale samples.
46
+
47
+ ## Dataset Details
48
+
49
+ The images are centered and normalized. They show a broad range of font styles.
50
+
51
+
52
+ ### Dataset Sources
53
+
54
+ - **Repository:** train set https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/multiclass/usps.bz2, test set: https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/multiclass/usps.t.bz2
55
+ - **Paper:** https://ieeexplore.ieee.org/abstract/document/291440
56
+
57
+ ## Uses
58
+
59
+ 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.
60
+
61
+ To partition the dataset, do the following.
62
+ 1. Install the package.
63
+ ```bash
64
+ pip install flwr-datasets[vision]
65
+ ```
66
+ 2. Use the HF Dataset under the hood in Flower Datasets.
67
+ ```python
68
+ from flwr_datasets import FederatedDataset
69
+ from flwr_datasets.partitioner import IidPartitioner
70
+
71
+ fds = FederatedDataset(
72
+ dataset="flwrlabs/usps",
73
+ partitioners={"train": IidPartitioner(num_partitions=10)}
74
+ )
75
+ partition = fds.load_partition(partition_id=0)
76
+ ```
77
+
78
+ ## Dataset Structure
79
+
80
+ ### Data Instances
81
+ The first instance of the train split is presented below:
82
+ ```
83
+ {
84
+ 'image': <PIL.PngImagePlugin.PngImageFile image mode=L size=16x16 at 0x133B4BA90>,
85
+ 'label': 6
86
+ }
87
+ ```
88
+ ### Data Split
89
+
90
+ ```
91
+ DatasetDict({
92
+ train: Dataset({
93
+ features: ['image', 'label'],
94
+ num_rows: 7291
95
+ })
96
+ test: Dataset({
97
+ features: ['image', 'label'],
98
+ num_rows: 2007
99
+ })
100
+ })
101
+ ```
102
+
103
+ ## Citation
104
+
105
+ When working with the USPS dataset, please cite the original paper.
106
+ If you're using this dataset with Flower Datasets and Flower, cite Flower.
107
+
108
+ **BibTeX:**
109
+
110
+ Original paper:
111
+ ```
112
+ @article{hull1994database,
113
+ title={A database for handwritten text recognition research},
114
+ journal={IEEE Transactions on pattern analysis and machine intelligence},
115
+ volume={16},
116
+ number={5},
117
+ pages={550--554},
118
+ year={1994},
119
+ publisher={IEEE}
120
+ }
121
+ ````
122
+
123
+ Flower:
124
+
125
+ ```
126
+ @article{DBLP:journals/corr/abs-2007-14390,
127
+ author = {Daniel J. Beutel and
128
+ Taner Topal and
129
+ Akhil Mathur and
130
+ Xinchi Qiu and
131
+ Titouan Parcollet and
132
+ Nicholas D. Lane},
133
+ title = {Flower: {A} Friendly Federated Learning Research Framework},
134
+ journal = {CoRR},
135
+ volume = {abs/2007.14390},
136
+ year = {2020},
137
+ url = {https://arxiv.org/abs/2007.14390},
138
+ eprinttype = {arXiv},
139
+ eprint = {2007.14390},
140
+ timestamp = {Mon, 03 Aug 2020 14:32:13 +0200},
141
+ biburl = {https://dblp.org/rec/journals/corr/abs-2007-14390.bib},
142
+ bibsource = {dblp computer science bibliography, https://dblp.org}
143
+ }
144
+ ```
145
+
146
+ ## Dataset Card Contact
147
+
148
+ In case of any doubts about the dataset preprocessing and preparation, please contact [Flower Labs](https://flower.ai/).