csaybar commited on
Commit
7213f00
·
verified ·
1 Parent(s): 3272ec0

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +132 -21
README.md CHANGED
@@ -11,65 +11,158 @@ tags:
11
  - remote-sensing
12
  pretty_name: cloudsen12plus
13
  ---
 
 
 
 
 
 
 
 
 
14
  # cloudsen12plus
15
 
16
  ****The largest dataset of expert-labeled pixels for cloud and cloud shadow detection in Sentinel-2****
17
 
18
 
19
- CloudSEN12+ is a significant extension of the CloudSEN12 dataset, which doubles the number of
20
  expert-reviewed labels, making it, by a large margin, the largest cloud detection dataset to
21
  date for Sentinel-2. All labels from the previous version have been curated and refined, enhancing
22
- the dataset's trustworthiness. This new release is licensed under CC0, which puts it in the public
23
  domain and allows anyone to use, modify, and distribute it without permission or attribution.
24
- The images have been padded from 509x509 to 512x512 and 2000x2000 to 2048x2048 to ensure that the
 
25
  patches are divisible by 32. The padding is filled with zeros in the left and bottom sides of the
26
  image. For those who prefer traditional storage formats, GeoTIFF files are available in our
27
  [ScienceDataBank](https://www.scidb.cn/en/detail?dataSetId=2036f4657b094edfbb099053d6024b08&version=V1)
28
  repository.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
29
  <center>
30
  <img src='https://cdn-uploads.huggingface.co/production/uploads/6402474cfa1acad600659e92/9UA4U3WObVeq7BAcf37-C.png' alt='drawing' width='50%'/>
31
  </center>
32
  *CloudSEN12+ spatial coverage. The terms p509 and p2000 denote the patch size 509 × 509 and 2000 × 2000,
33
  respectively. `high`, `scribble`, and `nolabel` refer to the types of expert-labeled annotations*
34
 
35
- ## 🌮 TACO Snippet
 
 
 
 
 
 
36
 
37
 
38
  Load this dataset using the `tacoreader` library.
 
39
  ```python
40
  import tacoreader
41
- dataset = tacoreader.load('...')
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
42
  ```
43
 
 
 
 
44
 
45
 
46
- Or in R:
47
- ```r
48
- library(tacoreader)
49
- dataset <- tacoreader::load('...')
50
- ```
51
  ## 🛰️ Sensor Information
52
 
53
-
54
  The sensor related to the dataset: **sentinel2msi**
55
- ## 🎯 Task
56
 
 
57
 
58
  The task associated with this dataset: **semantic-segmentation**
59
- ## 📂 Original Data Repository
60
 
 
61
 
62
  Source location of the raw data:**[https://huggingface.co/datasets/isp-uv-es/CloudSEN12Plus](https://huggingface.co/datasets/isp-uv-es/CloudSEN12Plus)**
63
- ## 💬 Discussion
64
 
 
65
 
66
  Insights or clarifications about the dataset: **[https://huggingface.co/datasets/tacofoundation/cloudsen12/discussions](https://huggingface.co/datasets/tacofoundation/cloudsen12/discussions)**
67
- ## 🔀 Split Strategy
68
 
 
69
 
70
  How the dataset is divided for training, validation, and testing: **stratified**
71
- ## 📚 Scientific Publications
72
 
 
73
 
74
  Publications that reference or describe the dataset.
75
 
@@ -90,9 +183,8 @@ Publications that reference or describe the dataset.
90
  }
91
  ```
92
 
93
-
94
-
95
  ### Publication 02
 
96
  - **DOI**: [10.1109/IGARSS52108.2023.10282381](10.1109/IGARSS52108.2023.10282381)
97
  - **Summary**: Exploration of incorrect annotations in cloud semantic segmentation datasets.
98
  - **BibTeX Citation**:
@@ -107,9 +199,8 @@ Publications that reference or describe the dataset.
107
  }
108
  ```
109
 
110
-
111
-
112
  ### Publication 03
 
113
  - **DOI**: [10.1016/j.dib.2024.110852](10.1016/j.dib.2024.110852)
114
  - **Summary**: Extended version of CloudSEN12. We include 2000 x 2000 patches to the dataset.
115
  - **BibTeX Citation**:
@@ -157,7 +248,8 @@ The dataset contains four classes: clear, thick cloud, thin cloud, and cloud sha
157
  ## 🌈 Optical Bands
158
 
159
 
160
- Spectral bands related to the sensor.
 
161
  |**Name**|**Common Name**|**Description**|**Center Wavelength**|**Full Width Half Max**|**Index**|
162
  | :--- | :--- | :--- | :--- | :--- | :--- |
163
  |B01|coastal aerosol|Band 1 - Coastal aerosol - 60m|443.5|17.0|0|
@@ -173,3 +265,22 @@ Spectral bands related to the sensor.
173
  |B10|cirrus|Band 10 - Cirrus - 60m|1375.5|31.0|10|
174
  |B11|SWIR 1|Band 11 - Shortwave infrared 1 - 20m|1613.5|89.0|11|
175
  |B12|SWIR 2|Band 12 - Shortwave infrared 2 - 20m|2199.5|173.0|12|
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
11
  - remote-sensing
12
  pretty_name: cloudsen12plus
13
  ---
14
+
15
+ <div style="text-align: center; border: 1px solid #ddd; border-radius: 10px; padding: 15px; max-width: 250px; margin: auto; background-color: #f9f9f9;">
16
+
17
+ ![Dataset Image](https://tacofoundation.github.io/taco.png)
18
+
19
+ <b><p>This dataset follows the TACO specification.</p></b>
20
+ </div>
21
+
22
+
23
  # cloudsen12plus
24
 
25
  ****The largest dataset of expert-labeled pixels for cloud and cloud shadow detection in Sentinel-2****
26
 
27
 
28
+ CloudSEN12+ version 1.1.0 is a significant extension of the CloudSEN12 dataset, which doubles the number of
29
  expert-reviewed labels, making it, by a large margin, the largest cloud detection dataset to
30
  date for Sentinel-2. All labels from the previous version have been curated and refined, enhancing
31
+ the dataset's truestworthiness. This new release is licensed under CC0, which puts it in the public
32
  domain and allows anyone to use, modify, and distribute it without permission or attribution.
33
+
34
+ The images are padded from 509x509 to 512x512 and 2000x2000 to 2048x2048 to ensure that the
35
  patches are divisible by 32. The padding is filled with zeros in the left and bottom sides of the
36
  image. For those who prefer traditional storage formats, GeoTIFF files are available in our
37
  [ScienceDataBank](https://www.scidb.cn/en/detail?dataSetId=2036f4657b094edfbb099053d6024b08&version=V1)
38
  repository.
39
+
40
+ `CloudSEN12+` v.1.1.0 offers three distinct modes, tailored for diverse research and application needs:
41
+
42
+ - **`cloudsen12-l1c`**: Patches derived from Sentinel-2 Level-1C imagery, including high-quality labels, scribble annotations, and unlabeled data.
43
+
44
+ - **`cloudsen12-l2a`**: Similar to cloudsen12-l1c but based on Sentinel-2 Level-2A data as processed by Google Earth Engine.
45
+
46
+ - **`cloudsen12-extra`**: A supplementary collection of metadata to enhance contextual understanding of landscapes. Cloud masks from multiple sources have been normalized to align with the CloudSEN12 class schema. This mode includes:
47
+ - **`s1_vv:`** Normalized Sentinel-1 Global Backscatter Model Land Surface (VV polarization).
48
+ - **`s1_vh:`** Normalized Sentinel-1 Global Backscatter Model Land Surface (VH polarization).
49
+ - **`elevation:`** Elevation data (meters) sourced from the MERIT Hydro dataset.
50
+ - **`LC10:`** ESA WorldCover 10m v100 land cover product.
51
+ - **`cloudmask_qa:`** Cloud mask from Sentinel-2 Level-1C.
52
+ - **`cloudmask_sen2cor:`** Cloud mask from Sentinel-2 Level-2A.
53
+ - **`cloudmask_cloudscore:`** Cloud detection results from Google Cloud Masking.
54
+ - **`cloudmask_s2cloudless:`** Cloud mask generated by Sentinel Hub Cloud Detector.
55
+ - **`cloudmask_unetmobv2:`** A cloud mask produced by a UnetMobV2 model trained on the CloudSEN12 dataset.
56
+
57
+ <center>
58
+ <img src='https://cdn-uploads.huggingface.co/production/uploads/6402474cfa1acad600659e92/9UA4U3WObVeq7BAcf37-C.png' alt='drawing' width='50%'/>
59
+ </center>
60
+
61
+ *CloudSEN12+ spatial coverage. The terms p509 and p2000 denote the patch size 509 × 509 and 2000 × 2000,
62
+ respectively. `high`, `scribble`, and `nolabel` refer to the types of expert-labeled annotations*
63
+
64
+ - Changelog:
65
+ Version 1.1.0:
66
+ - We save all GeoTIFF files with discard_lsb=2 to improve the compression ratio.
67
+ - Fixed 2000x2000 rotated patches. The datapoints are now correctly oriented. Check the patches:
68
+ - ROI_2526__20200709T105031_20200709T105719_T31UDQ
69
+ - ROI_0070__20190708T130251_20190708T130252_T24MUA
70
+ - ROI_4565__20200530T100029_20200530T100502_T32TQP
71
+ - Improved the quality of the following patches:
72
+ - ROI_1098__20200515T190909_20200515T191310_T11WPN
73
+ - ROI_1735__20190814T163849_20190814T164716_T15SXS
74
+ - ROI_0760__20190516T022551_20190516T022553_T56WMD
75
+ - ROI_3696__20200419T075611_20200419T080344_T35MRN
76
+ - ROI_2864__20170529T105621_20170529T110523_T31TCN
77
+ - We removed the following patches due to poor quality:
78
+ - ROI_3980__20190228T005641_20190228T005640_T58WDB
79
+ - ROI_1489__20210228T070831_20210228T070834_T40TDP
80
+
81
+ - Consideration:
82
+ - The field `roi_id` field serves as a unique identifier for the geographical location of each patch. In other words, it is used to link S2 images with
83
+ a specific geographic location. However, the roi_id between the 509x509 and 2000x2000 patches are
84
+ not the same. For example, the roid_id: `ROI_0008` in the 509x509 patches is not the same as the
85
+ `ROI_0008` in the 2000x2000 patches. In this version, we fixed this issue by summing the max value
86
+ of the 509x509 patches to the 2000x2000 patches. In this way, the `roi_id` between the 509x509 and
87
+ 2000x2000 patches are unique. If users of 2000x2000 patches need to match the original roi_id published
88
+ in the previous version, they can use the following formula:
89
+ - `old_roi_id_2000 = old_roi_id_2000 - 12101`
90
+ where `12101` is the max value of the 509 patches. We also reported the previous roi as old_roi_id.
91
+
92
  <center>
93
  <img src='https://cdn-uploads.huggingface.co/production/uploads/6402474cfa1acad600659e92/9UA4U3WObVeq7BAcf37-C.png' alt='drawing' width='50%'/>
94
  </center>
95
  *CloudSEN12+ spatial coverage. The terms p509 and p2000 denote the patch size 509 × 509 and 2000 × 2000,
96
  respectively. `high`, `scribble`, and `nolabel` refer to the types of expert-labeled annotations*
97
 
98
+
99
+
100
+ ## 🔄 Reproducible Example
101
+
102
+ <a target="_blank" href="https://colab.research.google.com/drive/1U9n40rwdnn73bdWruONA3hIs1-H3f74Q">
103
+ <img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/>
104
+ </a>
105
 
106
 
107
  Load this dataset using the `tacoreader` library.
108
+
109
  ```python
110
  import tacoreader
111
+ import rasterio as rio
112
+
113
+ print(tacoreader.__version__) # 0.5.2
114
+
115
+ # Remotely load the Cloud-Optimized Dataset
116
+ dataset = tacoreader.load("tacofoundation:cloudsen12-l1c")
117
+ #dataset = tacoreader.load("tacofoundation:cloudsen12-l2a")
118
+
119
+ # Read a sample
120
+ sample_idx = 2422
121
+ s2_l1c = dataset.read(sample_idx).read(0)
122
+ s2_label = dataset.read(sample_idx).read(1)
123
+
124
+ # Retrieve the data
125
+ with rio.open(s2_l1c) as src, rio.open(s2_label) as dst:
126
+ s2_l1c_data = src.read([4, 3, 2], window=rio.windows.Window(0, 0, 512, 512))
127
+ s2_label_data = dst.read(window=rio.windows.Window(0, 0, 512, 512))
128
+
129
+ # Display
130
+ fig, ax = plt.subplots(1, 2, figsize=(10, 5))
131
+ ax[0].imshow(s2_l1c_data.transpose(1, 2, 0) / 3000)
132
+ ax[0].set_title("Sentinel-2 L1C")
133
+ ax[1].imshow(s2_label_data[0])
134
+ ax[1].set_title("Human Label")
135
+ plt.tight_layout()
136
+ plt.savefig("taco_check.png")
137
+ plt.close(fig)
138
  ```
139
 
140
+ <center>
141
+ <img src='https://cdn-uploads.huggingface.co/production/uploads/6402474cfa1acad600659e92/0nRv7sqMRMNY-TVkY2kh7.png' alt='drawing' width='70%'/>
142
+ </center>
143
 
144
 
 
 
 
 
 
145
  ## 🛰️ Sensor Information
146
 
 
147
  The sensor related to the dataset: **sentinel2msi**
 
148
 
149
+ ## 🎯 Task
150
 
151
  The task associated with this dataset: **semantic-segmentation**
 
152
 
153
+ ## 📂 Original Data Repository
154
 
155
  Source location of the raw data:**[https://huggingface.co/datasets/isp-uv-es/CloudSEN12Plus](https://huggingface.co/datasets/isp-uv-es/CloudSEN12Plus)**
 
156
 
157
+ ## 💬 Discussion
158
 
159
  Insights or clarifications about the dataset: **[https://huggingface.co/datasets/tacofoundation/cloudsen12/discussions](https://huggingface.co/datasets/tacofoundation/cloudsen12/discussions)**
 
160
 
161
+ ## 🔀 Split Strategy
162
 
163
  How the dataset is divided for training, validation, and testing: **stratified**
 
164
 
165
+ ## 📚 Scientific Publications
166
 
167
  Publications that reference or describe the dataset.
168
 
 
183
  }
184
  ```
185
 
 
 
186
  ### Publication 02
187
+
188
  - **DOI**: [10.1109/IGARSS52108.2023.10282381](10.1109/IGARSS52108.2023.10282381)
189
  - **Summary**: Exploration of incorrect annotations in cloud semantic segmentation datasets.
190
  - **BibTeX Citation**:
 
199
  }
200
  ```
201
 
 
 
202
  ### Publication 03
203
+
204
  - **DOI**: [10.1016/j.dib.2024.110852](10.1016/j.dib.2024.110852)
205
  - **Summary**: Extended version of CloudSEN12. We include 2000 x 2000 patches to the dataset.
206
  - **BibTeX Citation**:
 
248
  ## 🌈 Optical Bands
249
 
250
 
251
+ Spectral bands related to the sensor: L1C
252
+
253
  |**Name**|**Common Name**|**Description**|**Center Wavelength**|**Full Width Half Max**|**Index**|
254
  | :--- | :--- | :--- | :--- | :--- | :--- |
255
  |B01|coastal aerosol|Band 1 - Coastal aerosol - 60m|443.5|17.0|0|
 
265
  |B10|cirrus|Band 10 - Cirrus - 60m|1375.5|31.0|10|
266
  |B11|SWIR 1|Band 11 - Shortwave infrared 1 - 20m|1613.5|89.0|11|
267
  |B12|SWIR 2|Band 12 - Shortwave infrared 2 - 20m|2199.5|173.0|12|
268
+
269
+ Spectral bands related to the sensor: L2A
270
+
271
+ | Band | Name | Description | Center Wavelength (nm) | Bandwidth (nm) | Index |
272
+ | :--- | :----------- | :------------------------------------------------------------------------------------------- | :--------------------- | :------------ | :---- |
273
+ | B01 | Coastal aerosol | Band 1 - Coastal aerosol - 60m | 443.5 | 17.0 | 0 |
274
+ | B02 | Blue | Band 2 - Blue - 10m | 496.5 | 53.0 | 1 |
275
+ | B03 | Green | Band 3 - Green - 10m | 560.0 | 34.0 | 2 |
276
+ | B04 | Red | Band 4 - Red - 10m | 664.5 | 29.0 | 3 |
277
+ | B05 | Red edge 1 | Band 5 - Vegetation red edge 1 - 20m | 704.5 | 13.0 | 4 |
278
+ | B06 | Red edge 2 | Band 6 - Vegetation red edge 2 - 20m | 740.5 | 13.0 | 5 |
279
+ | B07 | Red edge 3 | Band 7 - Vegetation red edge 3 - 20m | 783.0 | 18.0 | 6 |
280
+ | B08 | NIR | Band 8 - Near infrared - 10m | 840.0 | 114.0 | 7 |
281
+ | B8A | Red edge 4 | Band 8A - Vegetation red edge 4 - 20m | 864.5 | 19.0 | 8 |
282
+ | B09 | Water vapor | Band 9 - Water vapor - 60m | 945.0 | 18.0 | 9 |
283
+ | B11 | SWIR 1 | Band 11 - Shortwave infrared 1 - 20m | 1613.5 | 89.0 | 10 |
284
+ | B12 | SWIR 2 | Band 12 - Shortwave infrared 2 - 20m | 2199.5 | 173.0 | 11 |
285
+ | AOT | - | Aerosol Optical Thickness | - | - | 12 |
286
+ | WVP | - | Water Vapor Pressure. The height the water would occupy if the vapor were condensed into liquid and spread evenly across the column | - | - | 13 |