...
 
# BigEarthNet Deep Learning Models
This repository contains code to use the BigEarthNet archive for deep learning applications.
This repository contains code to use the BigEarthNet archive with the original CORINE Land Cover (CLC) Level-3 class nomenclature for deep learning applications.
If you use BigEarthNet archive, please cite our paper given below:
If you use the BigEarthNet archive or our pre-trained models, please cite the paper given below:
> G. Sumbul, M. Charfuelan, B. Demir, V. Markl, BigEarthNet: A Large-Scale Benchmark Archive for Remote Sensing Image Understanding, IEEE International Conference on Geoscience and Remote Sensing Symposium, pp. 5901-5904, Yokohama, Japan, 2019.
......@@ -17,10 +17,11 @@ If you use BigEarthNet archive, please cite our paper given below:
}
```
If you are interested in the new nomenclature of a smaller number of classes (BigEarthNet-19), check [here](https://gitlab.tubit.tu-berlin.de/rsim/bigearthnet-19-models).
# Pre-trained Deep Learning Models on BigEarthNet
We provide code and model weights for fhe following deep learning models that have been pre-trained on BigEarthNet for scene classification:
We provide code and model weights for the following deep learning models that have been pre-trained on BigEarthNet with the original Level-3 class nomenclature of CLC 2018 for scene classification:
Deep Learning Models pre-trained on the BigEarthNet with multi-labels associated to Level-3 class nomenclature of CLC 2018:
| Model Names | Pre-Trained TensorFlow Models | F<sub>1</sub> Score |
| ------------- |-------------------------------------------------------------------------------------------------------------------------------------------------------------|---------------------|
......@@ -31,7 +32,7 @@ Deep Learning Models pre-trained on the BigEarthNet with multi-labels associated
| ResNet101 | [http://bigearth.net/static/pretrained-models/original_labels/ResNet101.zip](http://bigearth.net/static/pretrained-models/original_labels/ResNet101.zip) | 72.76% |
| ResNet152 | [http://bigearth.net/static/pretrained-models/original_labels/ResNet152.zip](http://bigearth.net/static/pretrained-models/original_labels/ResNet152.zip) | 75.86% |
The results provided in the [BigEarthNet paper](http://bigearth.net/static/documents/BigEarthNet_IGARSS_2019.pdf) is different from those given above due to the selection of different train, validation and test sets.
The results provided in the [BigEarthNet paper](http://bigearth.net/static/documents/BigEarthNet_IGARSS_2019.pdf) are different from those given above due to the selection of different train, validation and test sets.
The TensorFlow code for these models can be found [here](https://gitlab.tu-berlin.de/rsim/bigearthnet-models-tf).
......@@ -42,10 +43,10 @@ After downloading the raw images from https://www.bigearth.net, they need to be
* `-o` or `--out_folder`: The output folder where the resulting files will be created.
* `-n` or `--splits`: A list of CSV files each of which contains the patch names of corresponding split.
To run the script, either GDAL or rasterio package should be installed. TensorFlow package should also be installed. The script is tested with Python 2.7, TensorFlow 1.3 and Ubuntu 16.04.
To run the script, either the GDAL or the rasterio package should be installed. The TensorFlow package should also be installed. The script is tested with Python 2.7, TensorFlow 1.3 and Ubuntu 16.04.
# Bugs and Requests
If you face a bug or have a feature request, please create an issue:
If you face a bug or have a feature request, please create an issue here:
https://gitlab.tubit.tu-berlin.de/rsim/bigearthnet-models/issues
Authors
......
......@@ -186,5 +186,3 @@ if __name__ == "__main__":
writer_list[split_idx]
)
writer_list[split_idx].close()