Code for paper "Context-self contrastive pretraining for crop type semantic segmentation"

Overview

Code for paper "Context-self contrastive pretraining for crop type semantic segmentation"

Setting up a python environment

  • Follow the instruction in https://docs.conda.io/projects/conda/en/latest/user-guide/install/linux.html for downloading and installing Miniconda

  • Open a terminal in the code directory

  • Create an environment using the .yml file:

    conda env create -f deepsatmodels_env.yml

  • Activate the environment:

    source activate deepsatmodels

  • Install required version of torch:

    conda install pytorch torchvision torchaudio cudatoolkit=10.1 -c pytorch-nightly

Datasets

MTLCC dataset (Germany)

Download the dataset (.tfrecords)

The data for Germany can be downloaded from: https://github.com/TUM-LMF/MTLCC

  • clone the repository in a separate directory:

    git clone https://github.com/TUM-LMF/MTLCC

  • move to the MTLCC root directory:

    cd MTLCC

  • download the data (40 Gb):

    bash download.sh full

Transform the dataset (.tfrecords -> .pkl)

  • go to the "CSCL_code" home directory:

    cd <.../CSCL_code>

  • activate the "cssl" python environment:

    conda activate cscl

  • add "CSCL_code" home directory to PYTHONPATH:

    export PYTHONPATH="<.../CSCL_code>:$PYTHONPATH"

  • Run the "data/MTLCC/make_pkl_dataset.py" script. Parameter numworkers defines the number of parallel processes employed:

    python data/MTLCC/make_pkl_dataset.py --rootdir <.../MTLCC> --numworkers

  • Running the above script will have the following effects:

    • will create a paths file for the tfrecords files in ".../MTLCC/data_IJGI18/datasets/full/tfrecords240_paths.csv"
    • will create a new directory to save data ".../MTLCC/data_IJGI18/datasets/full/240pkl"
    • will save data in ".../MTLCC/data_IJGI18/datasets/full/240pkl/ "
    • will save relative paths for all data, train data, eval data in ".../MTLCC/data_IJGI18/datasets/full/240pkl"

T31TFM_1618 dataset (France)

Download the dataset

The T31TFM_1618 dataset can be downloaded from Google drive here. Unzipping will create the following folder tree.

T31TFM_1618
├── 2016
│   ├── pkl_timeseries
│       ├── W799943_N6568107_E827372_S6540681
│       |   └── 6541426_800224_2016.pickle
|       |   └── ...
|       ├── ...
├── 2017
│   ├── pkl_timeseries
│       ├── W854602_N6650582_E882428_S6622759
│       |   └── 6623702_854602_2017.pickle
|       |   └── ...
|       ├── ...
├── 2018
│   ├── pkl_timeseries
│       ├── W882228_N6595532_E909657_S6568107
│       |   └── 6568846_888751_2018.pickle
|       |   └── ...
|       ├── ...
├── deepsatdata
|   └── T31TFM_16_products.csv
|   └── ...
|   └── T31TFM_16_parcels.csv
|   └── ...
└── paths
    └── train_paths.csv
    └── eval_paths.csv

Recreate the dataset from scratch

To recreate the dataset use the DeepSatData data generation pipeline.

  • Clone and move to the DeepSatData base directory
git clone https://github.com/michaeltrs/DeepSatData
cd .../DeepSatData
  • Download the Sentinel-2 products.
sh download/download.sh .../T31TFM_16_parcels.csv,.../T31TFM_17_parcels.csv,.../T31TFM_18_parcels.csv
  • Generate a labelled dataset (use case 1) for each year.
sh dataset/labelled_dense/make_labelled_dataset.sh ground_truths_file=<1:ground_truths_file> products_dir=<2:products_dir> labels_dir=<3:labels_dir> windows_dir=<4:windows_dir> timeseries_dir=<5:timeseries_dir> 
res=<6:res> sample_size=<7:sample_size> num_processes<8:num_processes> bands=<8:bands (optional)>

Experiments

Initial steps

  • Add the base directory and paths to train and evaluation path files in "data/datasets.yaml".

  • For each experiment we use a separate ".yaml" configuration file. Examples files are providedided in "configs". The default values filled in these files correspond to parameters used in the experiments presented in the paper.

  • activate "deepsatmodels" python environment:

    conda activate deepsatmodels

Model training

Modify respective .yaml config files accordingly to define the save directory or loading a pre-trained model from pre-trained checkpoints.

Randomly initialized "UNet3D" model

`python train_and_eval/segmentation_training.py --config_file configs/**/UNet3D.yaml --gpu_ids 0,1`

Randomly initialized "UNet2D-CLSTM" model

`python train_and_eval/segmentation_training.py --config_file configs/**/UNet2D_CLSTM.yaml --gpu_ids 0,1`

CSCL-pretrained "UNet2D-CLSTM" model

  • model pre-training

     python train_and_eval/segmentation_cscl_training.py --config_file configs/**/UNet2D_CLSTM_CSCL.yaml --gpu_ids 0,1
  • copy the path to the pre-training save directory in CHECKPOINT.load_from_checkpoint. This will load the latest saved model. To load a specific checkpoint copy the path to the .pth file

     python train_and_eval/segmentation_training.py --config_file configs/**/UNet2D_CLSTM.yaml --gpu_ids 0,1

Randomly initialized "UNet3Df" model

`python train_and_eval/segmentation_training.py --config_file configs/**/UNet3Df.yaml --gpu_ids 0,1`

CSCL-pretrained "UNet3Df" model

  • model pre-training

     python train_and_eval/segmentation_cscl_training.py --config_file configs/**/UNet3Df_CSCL.yaml --gpu_ids 0,1
  • copy the path to the pre-training save directory in CHECKPOINT.load_from_checkpoint. This will load the latest saved model. To load a specific checkpoint copy the path to the .pth file

     python train_and_eval/segmentation_training.py --config_file configs/**/UNet3Df.yaml --gpu_ids 0,1
Owner
Michael Tarasiou
Michael Tarasiou
DI-HPC is an acceleration operator component for general algorithm modules in reinforcement learning algorithms

DI-HPC: Decision Intelligence - High Performance Computation DI-HPC is an acceleration operator component for general algorithm modules in reinforceme

OpenDILab 185 Dec 29, 2022
Logistic Bandit experiments. Official code for the paper "Jointly Efficient and Optimal Algorithms for Logistic Bandits".

Code for the paper Jointly Efficient and Optimal Algorithms for Logistic Bandits, by Louis Faury, Marc Abeille, Clément Calauzènes and Kwang-Sun Jun.

Faury Louis 1 Jan 22, 2022
Tutoriais publicados nas nossas redes sociais para obtenção de dados, análises simples e outras tarefas relevantes no mercado financeiro.

Tutoriais Públicos Tutoriais publicados nas nossas redes sociais para obtenção de dados, análises simples e outras tarefas relevantes no mercado finan

Trading com Dados 68 Oct 15, 2022
MiniSom is a minimalistic implementation of the Self Organizing Maps

MiniSom Self Organizing Maps MiniSom is a minimalistic and Numpy based implementation of the Self Organizing Maps (SOM). SOM is a type of Artificial N

Giuseppe Vettigli 1.2k Jan 03, 2023
Project page for the paper Semi-Supervised Raw-to-Raw Mapping 2021.

Project page for the paper Semi-Supervised Raw-to-Raw Mapping 2021.

Mahmoud Afifi 22 Nov 08, 2022
NuPIC Studio is an all­-in-­one tool that allows users create a HTM neural network from scratch

NuPIC Studio is an all­-in-­one tool that allows users create a HTM neural network from scratch, train it, collect statistics, and share it among the members of the community. It is not just a visual

HTM Community 93 Sep 30, 2022
Official implement of Paper:A deeply supervised image fusion network for change detection in high resolution bi-temporal remote sening images

A deeply supervised image fusion network for change detection in high resolution bi-temporal remote sensing images 深度监督影像融合网络DSIFN用于高分辨率双时相遥感影像变化检测 Of

Chenxiao Zhang 135 Dec 19, 2022
A quantum game modeling of pandemic (QHack 2022)

Contributors: @JongheumJung, @YoonjaeChung, @GyunghunKim Abstract In the regime of a global pandemic, leaders around the world need to consider variou

Yoonjae Chung 8 Apr 03, 2022
A visualisation tool for Deep Reinforcement Learning

DRLVIS - Visualising Deep Reinforcement Learning Created by Marios Sirtmatsis with the support of Alex Bäuerle. DRLVis is an application used for visu

Marios Sirtmatsis 1 Nov 04, 2021
Pmapper is a super-resolution and deconvolution toolkit for python 3.6+

pmapper pmapper is a super-resolution and deconvolution toolkit for python 3.6+. PMAP stands for Poisson Maximum A-Posteriori, a highly flexible and a

NASA Jet Propulsion Laboratory 8 Nov 06, 2022
Code from Daniel Lemire, A Better Alternative to Piecewise Linear Time Series Segmentation

PiecewiseLinearTimeSeriesApproximation code from Daniel Lemire, A Better Alternative to Piecewise Linear Time Series Segmentation, SIAM Data Mining 20

Daniel Lemire 21 Oct 27, 2022
Ludwig is a toolbox that allows to train and evaluate deep learning models without the need to write code.

Translated in 🇰🇷 Korean/ Ludwig is a toolbox that allows users to train and test deep learning models without the need to write code. It is built on

Ludwig 8.7k Jan 05, 2023
Code for the ICCV2021 paper "Personalized Image Semantic Segmentation"

PSS: Personalized Image Semantic Segmentation Paper PSS: Personalized Image Semantic Segmentation Yu Zhang, Chang-Bin Zhang, Peng-Tao Jiang, Ming-Ming

张宇 15 Jul 09, 2022
This repository builds a basic vision transformer from scratch so that one beginner can understand the theory of vision transformer.

vision-transformer-from-scratch This repository includes several kinds of vision transformers from scratch so that one beginner can understand the the

1 Dec 24, 2021
🔥3D-RecGAN in Tensorflow (ICCV Workshops 2017)

3D Object Reconstruction from a Single Depth View with Adversarial Learning Bo Yang, Hongkai Wen, Sen Wang, Ronald Clark, Andrew Markham, Niki Trigoni

Bo Yang 125 Nov 26, 2022
A containerized REST API around OpenAI's CLIP model.

OpenAI's CLIP — REST API This is a container wrapping OpenAI's CLIP model in a RESTful interface. Running the container locally First, build the conta

Santiago Valdarrama 48 Nov 06, 2022
Wandb-predictions - WANDB Predictions With Python

WANDB API CI/CD Below we capture the CI/CD scenarios that we would expect with o

Anish Shah 6 Oct 07, 2022
Code for "Learning From Multiple Experts: Self-paced Knowledge Distillation for Long-tailed Classification", ECCV 2020 Spotlight

Learning From Multiple Experts: Self-paced Knowledge Distillation for Long-tailed Classification Implementation of "Learning From Multiple Experts: Se

27 Nov 05, 2022
Source-to-Source Debuggable Derivatives in Pure Python

Tangent Tangent is a new, free, and open-source Python library for automatic differentiation. Existing libraries implement automatic differentiation b

Google 2.2k Jan 01, 2023
Artifacts for paper "MMO: Meta Multi-Objectivization for Software Configuration Tuning"

MMO: Meta Multi-Objectivization for Software Configuration Tuning This repository contains the data and code for the following paper that is currently

0 Nov 17, 2021