Datasets:
Update README.md
Browse files
README.md
CHANGED
@@ -31,3 +31,93 @@ dataset_info:
|
|
31 |
download_size: 93987179
|
32 |
dataset_size: 87998673.69999999
|
33 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
31 |
download_size: 93987179
|
32 |
dataset_size: 87998673.69999999
|
33 |
---
|
34 |
+
|
35 |
+
# Dataset Card for EuroSat
|
36 |
+
|
37 |
+
## Table of Contents
|
38 |
+
- [How to Use](#How-to-Use)
|
39 |
+
- [Dataset Description](#dataset-description)
|
40 |
+
- [Dataset Summary](#dataset-summary)
|
41 |
+
- [Dataset Structure](#dataset-structure)
|
42 |
+
- [Data Instances](#data-instances)
|
43 |
+
- [Data Fields](#data-fields)
|
44 |
+
- [Data Splits](#data-splits)
|
45 |
+
- [Additional Information](#additional-information)
|
46 |
+
- [Licensing Information](#licensing-information)
|
47 |
+
|
48 |
+
|
49 |
+
## How to Use
|
50 |
+
|
51 |
+
- Install [datasets](https://pypi.org/project/datasets/):
|
52 |
+
|
53 |
+
```bash
|
54 |
+
pip install datasets
|
55 |
+
```
|
56 |
+
- How to use in Python
|
57 |
+
```py
|
58 |
+
from datasets import load_dataset
|
59 |
+
train_data = load_dataset("Honaker/eurosat_dataset", split="train")
|
60 |
+
```
|
61 |
+
|
62 |
+
## Dataset Description
|
63 |
+
- **Homepage:** https://zenodo.org/record/7711810#.ZAm3k-zMKEA
|
64 |
+
|
65 |
+
### Dataset Summary
|
66 |
+
EuroSat is an image classification dataset with 10 different classes on satellite imagery. There is over 27,000 labeled images.
|
67 |
+
|
68 |
+
|
69 |
+
## Dataset Structure
|
70 |
+
|
71 |
+
The dataset is structured as follows:
|
72 |
+
```py
|
73 |
+
DatasetDict({
|
74 |
+
train: Dataset({
|
75 |
+
features: ['image', 'labels'],
|
76 |
+
num_rows: 21600
|
77 |
+
})
|
78 |
+
validation: Dataset({
|
79 |
+
features: ['image', 'labels'],
|
80 |
+
num_rows: 2700
|
81 |
+
})
|
82 |
+
test: Dataset({
|
83 |
+
features: ['image', 'labels'],
|
84 |
+
num_rows: 2700
|
85 |
+
})
|
86 |
+
})
|
87 |
+
```
|
88 |
+
|
89 |
+
### Data Instances
|
90 |
+
|
91 |
+
An example of the data for one image is:
|
92 |
+
```py
|
93 |
+
{
|
94 |
+
'image': <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=64x64>,
|
95 |
+
'labels': 0
|
96 |
+
}
|
97 |
+
```
|
98 |
+
|
99 |
+
With the type of each field being defined as:
|
100 |
+
```py
|
101 |
+
{
|
102 |
+
'image': <PIL.JpegImagePlugin.JpegImageFile>,
|
103 |
+
'labels': Integer
|
104 |
+
```
|
105 |
+
### Data Fields
|
106 |
+
|
107 |
+
The dataset has the following fields:
|
108 |
+
|
109 |
+
- 'image': Satellite image that is of type <PIL.TiffImagePlugin.TiffImageFile image>
|
110 |
+
- 'labels': the label of the Satellite image as an integer
|
111 |
+
|
112 |
+
### Data Splits
|
113 |
+
|
114 |
+
| | Train | Validation | Test |
|
115 |
+
|----------------|--------|------------|------|
|
116 |
+
| Images | 21600 | 2700 | 2700 |
|
117 |
+
|
118 |
+
|
119 |
+
## Additional Information
|
120 |
+
|
121 |
+
### Licensing Information
|
122 |
+
EuroSat is licensed under a MIT
|
123 |
+
|