Large Scale High-Resolution Land Cover Mapping with Multi-Resolution Data

Repo accompanying the CVPR 2019 paper "Large Scale High-Resolution Land Cover Mapping with Multi-Resolution Data".

Data

The dataset and its documentation can be found at LILA.science. The download_all.sh script will download and unzip a copy of this dataset to the ./chesapeake_data/ directory. Note: this script assumes that the azcopy program is on your PATH; azcopy can be downloaded here.

Requirements

Pre-trained keras models were generated with Python 3.6. See requirements.txt for the library versions that we used when developing. Generally, these libraries are needed:

numpy
pandas
tensorflow-gpu
Keras
segmentation-models
shapely
rasterio

Install azcopy

wget https://aka.ms/downloadazcopy-v10-linux
tar -xvf downloadazcopy-v10-linux
sudo cp ./azcopy_linux_amd64_*/azcopy /usr/bin/

Only works with sudo azcopy.

Usage

Make use of run.py to run an end to end process (training, testing and evaluation):

nohup python run.py --name example_config

When leaving all arguments to default, they will be fetched from landcover/config.py.

If you'd like to overwrite the defaults, you can set the arguments in the command line:

nohup python run.py -v 3 --data-dir chesapeake_data/ --output-dir results/ \
--training-states ny_1m_2013 --validation-states ny_1m_2013 \
--test-states ny_1m_2013  --name example --model unet --batch-size 16

Or edit landcover/config.py.

Note: the nohup command ensures the process keeps running even if the terminal is disconnected. It will also save an output with the logs in a nohup.txt file.

TensorBoard

To display metrics in TensorBoard run in the instance:

gsutil rsync -r land-cover/logs/ gs://lc-training-data/logs/

Locally run:

tensorboard --logdir gs://lc-training-data/logs

Test

Run all tests:

pytest --pylint --cov=landcover/

Format code:

black

References

Please cite the following papers if you use this work:

@inproceedings{robinson2019large,
  title={Large Scale High-Resolution Land Cover Mapping With Multi-Resolution Data},
  author={Robinson, Caleb and Hou, Le and Malkin, Kolya and Soobitsky, Rachel and Czawlytko, Jacob and Dilkina, Bistra and Jojic, Nebojsa},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2019},
  url={http://openaccess.thecvf.com/content_CVPR_2019/html/Robinson_Large_Scale_High-Resolution_Land_Cover_Mapping_With_Multi-Resolution_Data_CVPR_2019_paper.html}
}

@inproceedings{malkin2018label,
  title={Label super-resolution networks},
  author={Malkin, Kolya and Robinson, Caleb and Hou, Le and Soobitsky, Rachel and Czawlytko, Jacob and Samaras, Dimitris and Saltz, Joel and Joppa, Lucas and Jojic, Nebojsa},
  booktitle={International Conference on Learning Representations (ICLR)},
  year={2019},
  url={https://openreview.net/forum?id=rkxwShA9Ym},
}

Todo

Change train_model_landcover.py to save models without the superres loss, jaccard loss, or Lambda layers as these can cause problems with saving/loading in different versions.
Create a test script that computes accuracy on the fly without saving model results.

calebrob6 / land-cover