Hugging Face's logo

Spaces:
BIFOLD-BigEarthNetv2-0
/
README
Running

App Files Community
1
hackelle commited on
Commit
f46dfb3
•
1 Parent(s): 732d001

Update README.md

Browse files

Adding information on new band order

Files changed (1) hide show
  1. README.md +31 -38
README.md CHANGED
@@ -6,26 +6,33 @@ colorTo: green
6
  sdk: static
7
  pinned: false
8
  license: mit
9
- short_description: Repository of Pretrained Model Weights on BigEarthNet v2.0
10
  ---
11
 
12
  [TU Berlin](https://www.tu.berlin/) | [RSiM](https://rsim.berlin/) | [DIMA](https://www.dima.tu-berlin.de/menue/database_systems_and_information_management_group/) | [BigEarth](http://www.bigearth.eu/) | [BIFOLD](https://bifold.berlin/)
13
  :---:|:---:|:---:|:---:|:---:
14
  <a href="https://www.tu.berlin/"><img src="https://raw.githubusercontent.com/wiki/lhackel-tub/ConfigILM/static/imgs/tu-berlin-logo-long-red.svg" style="font-size: 1rem; height: 2em; width: auto" alt="TU Berlin Logo"/> | <a href="https://rsim.berlin/"><img src="https://raw.githubusercontent.com/wiki/lhackel-tub/ConfigILM/static/imgs/RSiM_Logo_1.png" style="font-size: 1rem; height: 2em; width: auto" alt="RSiM Logo"> | <a href="https://www.dima.tu-berlin.de/menue/database_systems_and_information_management_group/"><img src="https://raw.githubusercontent.com/wiki/lhackel-tub/ConfigILM/static/imgs/DIMA.png" style="font-size: 1rem; height: 2em; width: auto" alt="DIMA Logo"> | <a href="http://www.bigearth.eu/"><img src="https://raw.githubusercontent.com/wiki/lhackel-tub/ConfigILM/static/imgs/BigEarth.png" style="font-size: 1rem; height: 2em; width: auto" alt="BigEarth Logo"> | <a href="https://bifold.berlin/"><img src="https://raw.githubusercontent.com/wiki/lhackel-tub/ConfigILM/static/imgs/BIFOLD_Logo_farbig.png" style="font-size: 1rem; height: 2em; width: auto; margin-right: 1em" alt="BIFOLD Logo">
15
 
16
- # BigEarthNet v2.0 Pretrained Model Weights
17
- We provide weights for several different pretrained models.
18
- The model weights for the best-performing model, based on the macro average precision score on the recommended test split, have been uploaded.
19
- All models have been trained using: i) BigEarthNet-S1 data only (S1), ii) BigEarthNet-S2 data only (S2), or iii) both BigEarthNet-S1 and -S2 (S1+S2) together.
20
 
21
- The following bands were used to train the models:
22
- - For models using BigEarthNet-S1 only: Sentinel-1 bands `["VH", "VV"]`
23
- - For models using BigEarthNet-S2 only: Sentinel-2 10m bands and 20m bands `["B02", "B03", "B04", "B08", "B05", "B06", "B07", "B11", "B12", "B8A"]`
24
- - For models using BigEarthNet-S1 and -S2: Sentinel-2 10m bands and 20m bands and Sentinel-1 bands = `["B02", "B03", "B04", "B08", "B05", "B06", "B07", "B11", "B12", "B8A", "VH", "VV"]`
 
 
 
 
 
 
 
25
 
26
 
27
  The multi-hot encoded output of the model indicates the predicted multi-label output.
28
- The multi-hot encoded output relates to the following class labels sorted in alphabetical order:
29
  `['Agro-forestry areas', 'Arable land', 'Beaches, dunes, sands', 'Broad-leaved forest', 'Coastal wetlands',
30
  'Complex cultivation patterns', 'Coniferous forest', 'Industrial or commercial units', 'Inland waters',
31
  'Inland wetlands', 'Land principally occupied by agriculture, with significant areas of natural vegetation',
@@ -33,37 +40,31 @@ The multi-hot encoded output relates to the following class labels sorted in alp
33
  'Natural grassland and sparsely vegetated areas', 'Pastures', 'Permanent crops', 'Transitional woodland, shrub',
34
  'Urban fabric']`
35
 
36
-
37
  ![[BigEarthNet](http://bigearth.net/)](https://raw.githubusercontent.com/wiki/lhackel-tub/ConfigILM/static/imgs/combined_2000_600_2020_0_wide.jpg)
38
 
39
- ## Links
40
-
41
- | Model | Equivalent [`timm`](https://huggingface.co/docs/timm/en/index) model name | S1 only | S2 only | S1+S2 |
42
- |:-----------------|:---------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------:|:-----------------------------------------------------------------------------------------------:|:---------------------------------------------------------------------------------------------------:|
43
- | ConvMixer-768/32 | `convmixer_768_32` | [ConvMixer-768/32 S1](https://huggingface.co/BIFOLD-BigEarthNetv2-0/convmixer_768_32-s1-v0.1.1) | [ConvMixer-768/32 S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/convmixer_768_32-s2-v0.1.1) | [ConvMixer-768/32 S1+S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/convmixer_768_32-all-v0.1.1) |
44
- | ConvNext v2 Base | `convnextv2_base` | [ConvNext v2 Base S1](https://huggingface.co/BIFOLD-BigEarthNetv2-0/convnextv2_base-s1-v0.1.1) | [ConvNext v2 Base S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/convnextv2_base-s2-v0.1.1) | [ConvNext v2 Base S1+S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/convnextv2_base-all-v0.1.1) |
45
- | MLP-Mixer Base | `mixer_b16_224` | [MLP-Mixer Base S1](https://huggingface.co/BIFOLD-BigEarthNetv2-0/mixer_b16_224-s1-v0.1.1) | [MLP-Mixer Base S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/mixer_b16_224-s2-v0.1.1) | [MLP-Mixer Base S1+S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/mixer_b16_224-all-v0.1.1) |
46
- | MobileViT-S | `mobilevit_s` | [MobileViT-S S1](https://huggingface.co/BIFOLD-BigEarthNetv2-0/mobilevit_s-s1-v0.1.1) | [MobileViT-S S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/mobilevit_s-s2-v0.1.1) | [MobileViT-S S1+S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/mobilevit_s-all-v0.1.1) |
47
- | ResNet-50 | `resnet50` | [ResNet-50 S1](https://huggingface.co/BIFOLD-BigEarthNetv2-0/resnet50-s1-v0.1.1) | [ResNet-50 S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/resnet50-s2-v0.1.1) | [ResNet-50 S1+S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/resnet50-all-v0.1.1) |
48
- | ResNet-101 | `resnet101` | [ResNet-101 S1](https://huggingface.co/BIFOLD-BigEarthNetv2-0/resnet101-s1-v0.1.1) | [ResNet-101 S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/resnet101-s2-v0.1.1) | [ResNet-101 S1+S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/resnet101-all-v0.1.1) |
49
- | ViT Base | `vit_base_patch8_224` | [ViT Base S1](https://huggingface.co/BIFOLD-BigEarthNetv2-0/vit_base_patch8_224-s1-v0.1.1) | [ViT Base S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/vit_base_patch8_224-s2-v0.1.1) | [ViT Base S1+S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/vit_base_patch8_224-all-v0.1.1) |
50
 
51
 
52
  ![[BigEarthNet](http://bigearth.net/)](https://raw.githubusercontent.com/wiki/lhackel-tub/ConfigILM/static/imgs/combined_2000_600_2020_0_wide.jpg)
53
 
54
- ## Usage
55
-
56
  To use the model, download the codes that define the model architecture from the
57
- [official BigEarthNet v2.0 (reBEN) repository](https://git.tu-berlin.de/rsim/reben-training-scripts) and load the model with the corresponding weights
58
  using the code below. Note that [`configilm`](https://pypi.org/project/configilm/) is a requirement to use the
59
  code below.
60
 
61
  ```python
62
  from reben_publication.BigEarthNetv2_0_ImageClassifier import BigEarthNetv2_0_ImageClassifier
63
 
64
- model = BigEarthNetv2_0_ImageClassifier.from_pretrained(
65
- "path_to/huggingface_model_folder"
66
- )
67
  ```
68
 
69
  e.g.
@@ -72,21 +73,13 @@ e.g.
72
  from reben_publication.BigEarthNetv2_0_ImageClassifier import BigEarthNetv2_0_ImageClassifier
73
 
74
  model = BigEarthNetv2_0_ImageClassifier.from_pretrained(
75
- "BIFOLD-BigEarthNetv2-0/resnet50-s2-v0.1.1"
76
  )
77
  ```
78
 
79
  If you use any of these models in your research, please cite the following papers:
80
  ```bibtex
81
- @article{clasen2024refinedbigearthnet,
82
- title={reBEN: Refined BigEarthNet Dataset for Remote Sensing Image Analysis},
83
- author={Clasen, Kai Norman and Hackel, Leonard and Burgert, Tom and Sumbul, Gencer and Demir, Beg{\"u}m and Markl, Volker},
84
- year={2024},
85
- eprint={2407.03653},
86
- archivePrefix={arXiv},
87
- primaryClass={cs.CV},
88
- url={https://arxiv.org/abs/2407.03653},
89
- }
90
  ```
91
  ```bibtex
92
  @article{hackel2024configilm,
 
6
  sdk: static
7
  pinned: false
8
  license: mit
9
+ short_description: Official Repository of Pretrained Models on BigEarthNet v2.0
10
  ---
11
 
12
  [TU Berlin](https://www.tu.berlin/) | [RSiM](https://rsim.berlin/) | [DIMA](https://www.dima.tu-berlin.de/menue/database_systems_and_information_management_group/) | [BigEarth](http://www.bigearth.eu/) | [BIFOLD](https://bifold.berlin/)
13
  :---:|:---:|:---:|:---:|:---:
14
  <a href="https://www.tu.berlin/"><img src="https://raw.githubusercontent.com/wiki/lhackel-tub/ConfigILM/static/imgs/tu-berlin-logo-long-red.svg" style="font-size: 1rem; height: 2em; width: auto" alt="TU Berlin Logo"/> | <a href="https://rsim.berlin/"><img src="https://raw.githubusercontent.com/wiki/lhackel-tub/ConfigILM/static/imgs/RSiM_Logo_1.png" style="font-size: 1rem; height: 2em; width: auto" alt="RSiM Logo"> | <a href="https://www.dima.tu-berlin.de/menue/database_systems_and_information_management_group/"><img src="https://raw.githubusercontent.com/wiki/lhackel-tub/ConfigILM/static/imgs/DIMA.png" style="font-size: 1rem; height: 2em; width: auto" alt="DIMA Logo"> | <a href="http://www.bigearth.eu/"><img src="https://raw.githubusercontent.com/wiki/lhackel-tub/ConfigILM/static/imgs/BigEarth.png" style="font-size: 1rem; height: 2em; width: auto" alt="BigEarth Logo"> | <a href="https://bifold.berlin/"><img src="https://raw.githubusercontent.com/wiki/lhackel-tub/ConfigILM/static/imgs/BIFOLD_Logo_farbig.png" style="font-size: 1rem; height: 2em; width: auto; margin-right: 1em" alt="BIFOLD Logo">
15
 
16
+ # BigEarthNet v2.0 Pretrained Weights
17
+ We provide pretrained weights for several different models.
18
+ The weights for the best-performing model, based on the Macro Average Precision score on the recommended test split, have been uploaded.
19
+ All models have been trained using: i) Sentinel-1 data only (S1), ii) Sentinel-2 data only (S2), or iii) both Sentinel-1 and Sentinel-2 (S1+S2) modalities together.
20
 
21
+ The following bands (in the specified order) were used to train the models with version 0.2.0:
22
+ - For models using Sentinel-1 only: Sentinel-1 bands `["VV", "VH"]`
23
+ - For models using Sentinel-2 only: Sentinel-2 10m bands and 20m bands `["B02", "B03", "B04", "B05", "B06", "B07", "B08", "B8A", "B11", "B12"]`
24
+ - For models using Sentinel-1 and Sentinel-2: Sentinel-1 bands and Sentinel-2 10m bands and 20m bands `["VV", "VH", "B02", "B03", "B04", "B05", "B06", "B07", "B08", "B8A", "B11", "B12"]`
25
+
26
+ > **_NOTE:_** Older versions of the models have been trained with different band orders that are not compatible with the current version and do not match the order proposed in the technical documentation of Sentinel-2.
27
+ >
28
+ > The following bands (in the specified order) were used to train the models with version 0.1.1:
29
+ > - For models using Sentinel-1 only: Sentinel-1 bands `["VH", "VV"]`
30
+ > - For models using Sentinel-2 only: Sentinel-2 10m bands and 20m bands `["B02", "B03", "B04", "B08", "B05", "B06", "B07", "B11", "B12", "B8A"]`
31
+ > - For models using Sentinel-1 and Sentinel-2: Sentinel-2 10m bands and 20m bands and Sentinel-1 bands = `["B02", "B03", "B04", "B08", "B05", "B06", "B07", "B11", "B12", "B8A", "VH", "VV"]`
32
 
33
 
34
  The multi-hot encoded output of the model indicates the predicted multi-label output.
35
+ The multi-hot encoded output relates to the following classes sorted in alphabetical order:
36
  `['Agro-forestry areas', 'Arable land', 'Beaches, dunes, sands', 'Broad-leaved forest', 'Coastal wetlands',
37
  'Complex cultivation patterns', 'Coniferous forest', 'Industrial or commercial units', 'Inland waters',
38
  'Inland wetlands', 'Land principally occupied by agriculture, with significant areas of natural vegetation',
 
40
  'Natural grassland and sparsely vegetated areas', 'Pastures', 'Permanent crops', 'Transitional woodland, shrub',
41
  'Urban fabric']`
42
 
43
+ ## <!-- Tabelle -->
44
  ![[BigEarthNet](http://bigearth.net/)](https://raw.githubusercontent.com/wiki/lhackel-tub/ConfigILM/static/imgs/combined_2000_600_2020_0_wide.jpg)
45
 
46
+ | Model | Equivalent [`timm`](https://huggingface.co/docs/timm/en/index) model name | Sentinel-1 only | Sentinel-2 only | Sentinel-1 and Sentinel-2 |
47
+ |:-----------------|:---------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------:|:-----------------------------------------------------------------------------------------------:|:------------------------------------------------------------------------------------------------:|
48
+ | ConvMixer-768/32 | `convmixer_768_32` | [ConvMixer-768/32 S1](https://huggingface.co/BIFOLD-BigEarthNetv2-0/convmixer_768_32-s1-v0.2.0) | [ConvMixer-768/32 S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/convmixer_768_32-s2-v0.2.0) | [ConvMixer-768/32 S1+S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/convmixer_768_32-all-v0.2.0) |
49
+ | ConvNext v2 Base | `convnextv2_base` | [ConvNext v2 Base S1](https://huggingface.co/BIFOLD-BigEarthNetv2-0/convnextv2_base-s1-v0.2.0) | [ConvNext v2 Base S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/convnextv2_base-s2-v0.2.0) | [ConvNext v2 Base S1+S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/convnextv2_base-all-v0.2.0) |
50
+ | MLP-Mixer Base | `mixer_b16_224` | [MLP-Mixer Base S1](https://huggingface.co/BIFOLD-BigEarthNetv2-0/mixer_b16_224-s1-v0.2.0) | [MLP-Mixer Base S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/mixer_b16_224-s2-v0.2.0) | [MLP-Mixer Base S1+S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/mixer_b16_224-all-v0.2.0) |
51
+ | MobileViT-S | `mobilevit_s` | [MobileViT-S S1](https://huggingface.co/BIFOLD-BigEarthNetv2-0/mobilevit_s-s1-v0.2.0) | [MobileViT-S S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/mobilevit_s-s2-v0.2.0) | [MobileViT-S S1+S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/mobilevit_s-all-v0.2.0) |
52
+ | ResNet-50 | `resnet50` | [ResNet-50 S1](https://huggingface.co/BIFOLD-BigEarthNetv2-0/resnet50-s1-v0.2.0) | [ResNet-50 S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/resnet50-s2-v0.2.0) | [ResNet-50 S1+S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/resnet50-all-v0.2.0) |
53
+ | ResNet-101 | `resnet101` | [ResNet-101 S1](https://huggingface.co/BIFOLD-BigEarthNetv2-0/resnet101-s1-v0.2.0) | [ResNet-101 S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/resnet101-s2-v0.2.0) | [ResNet-101 S1+S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/resnet101-all-v0.2.0) |
54
+ | ViT Base | `vit_base_patch8_224` | [ViT Base S1](https://huggingface.co/BIFOLD-BigEarthNetv2-0/vit_base_patch8_224-s1-v0.2.0) | [ViT Base S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/vit_base_patch8_224-s2-v0.2.0) | [ViT Base S1+S2](https://huggingface.co/BIFOLD-BigEarthNetv2-0/vit_base_patch8_224-all-v0.2.0) |
 
 
55
 
56
 
57
  ![[BigEarthNet](http://bigearth.net/)](https://raw.githubusercontent.com/wiki/lhackel-tub/ConfigILM/static/imgs/combined_2000_600_2020_0_wide.jpg)
58
 
 
 
59
  To use the model, download the codes that define the model architecture from the
60
+ [official BigEarthNet v2.0 (reBEN) repository](https://git.tu-berlin.de/rsim/reben-training-scripts) and load the model
61
  using the code below. Note that [`configilm`](https://pypi.org/project/configilm/) is a requirement to use the
62
  code below.
63
 
64
  ```python
65
  from reben_publication.BigEarthNetv2_0_ImageClassifier import BigEarthNetv2_0_ImageClassifier
66
 
67
+ model = BigEarthNetv2_0_ImageClassifier.from_pretrained("path_to/huggingface_model_folder")
 
 
68
  ```
69
 
70
  e.g.
 
73
  from reben_publication.BigEarthNetv2_0_ImageClassifier import BigEarthNetv2_0_ImageClassifier
74
 
75
  model = BigEarthNetv2_0_ImageClassifier.from_pretrained(
76
+ "BIFOLD-BigEarthNetv2-0/resnet50-s2-v0.2.0"
77
  )
78
  ```
79
 
80
  If you use any of these models in your research, please cite the following papers:
81
  ```bibtex
82
+ CITATION FOR DATASET PAPER
 
 
 
 
 
 
 
 
83
  ```
84
  ```bibtex
85
  @article{hackel2024configilm,