Scoring Frugality for Remote Sensing Data Processing Algorithms

This repository contains the code associated to the paper Scoring Frugality for Remote Sensing Data Processing Algorithms to carry out frugality analysis on clustering and change detection tasks.

In remote sensing, the demand for extensive data processing algorithms continues to grow exponentially. Assessing the frugality of data processing algorithms has become a priority in the machine learning and artificial intelligence community. However, defining a quantifiable measure that combines performance and energy consumption remains a complex challenge. This paper introduces a methodology for collecting energy consumption data and presents three different frugality scoring methods. Through a case study of two classical data processing tasks in remote sensing, change detection and clustering, we demonstrate that these three scores capture different aspects of frugality. We further suggest a combined approach for users to achieve a more comprehensive assessment.

This work rely on the submodule performance-tracking containing scripts for each data processing algorithms along with a guide and configuration files for the data collection pipeline.

WORK IN PROGRESS

Repository structure

.
├── data
│   └── ...
├── doc
│   ├── static
│   │   ├── Scene_1small.png
│   │   └── Scene_1small_truth.png
│   └── submit_EUSIPCO_2025.pdf
├── environment.yml
├── LICENCE
├── README.md
└── src
    └── performance-tracking
        ├── experiments
        │   ├── conso
        │   │   ├── analyse_stats.py
        │   │   ├── get_conso.py
        │   │   ├── get_stats.py
        │   │   ├── query_influx.sh
        │   │   ├── simulation_metrics_exec.sh
        │   │   ├── stats_summary_blob.py
        │   │   ├── stats_summary_deep.py
        │   │   └── stats_summary.py
        │   ├── conso_change
        │   │   ├── cd_sklearn_pair_var.py
        │   │   ├── change-detection.py
        │   │   ├── functions.py
        │   │   ├── get_perf.py
        │   │   ├── helpers
        │   │   │   └── multivariate_images_tool.py
        │   │   └── main.py
        │   ├── conso_classif_deep
        │   │   ├── classif_deep.py
        │   │   ├── get_perf.py
        │   │   ├── get_scores.py
        │   │   ├── read_event.py
        │   │   ├── read_events.py
        │   │   └── simulation_metrics_exec.sh
        │   └── conso_clustering
        │       ├── clustering_blob.py
        │       ├── clustering.py
        │       ├── get_perf_blob.py
        │       ├── get_perf.py
        │       ├── helpers
        │       │   └── processing_helpers.py
        │       ├── plot_clustering.py
        │       ├── utils_clustering_blob.py
        │       └── utils_clustering.py
        ├── plot_usage.py
        ├── README.md
        └── simulation_metrics_exec.sh

Installation

Dependencies

python 3.11.8, codecarbon 2.3.4, numpy 1.26.4, pandas 2.2.1, scikit-learn 1.4.1

conda env create -f environment.yml
conda activate frugal-score

Data

This repository presents the code for frugality evaluation for two remote sensing data processing tasks : clustering and change detection.

The clustering task was carried out on a toy dataset containing 1000 instances divided in 5 blobs on 5 features wuth a standard variation of 5. Its projection over the first 3 dimensions of the PCA is presented following

The change detection task was carried out on multiband PolSAR image time series from the UAVSAR (Courtesy NASA/JPL-Caltech) [1] with ground truth data from [2].

Usage

File	Associated command	Description
change-detection.py	python change-detection.py --storage_path [PATH_TO_FOLDER_TO_STORE_RESULTS] --image [PATH_TO_FOLDER_WITH_IMAGES] --window [WINDOW_SIZE] --cores [NUMBER_OF_CORES_USED] --number_run [NUMBER_OF_RUNS] --robust [ROBUSTNESS ID]	Runs change detection algorithms on UAVSAR data
clustering_blob.py	python clustering_blob.py --storage_path [PATH_TO_FOLDER_TO_STORE_RESULTS] --data_seed [SEED] --random_seed [SEED] --n_clusters [NUMBER_OF_CLUSTERS] --model [CLUSTERING_METHOD] --repeat [NUMBER_OF_MODEL_RUNS] --number_run/-n [NUMBER_OF_RUNS]	Runs clustering algorithms on toy data
plot_change_detection.py	python plot_change_detection.py --result_path [PATH_TO_RESULTS] --output_path [OUTPUT_PATH]	Plot score results for change detection
plot_clustering.py	python plot_clustering.py --result_path [PATH_TO_RESULTS] --output_path [OUTPUT_PATH]	Plot score results for clustering

Authors

• Matthieu Verlynde (matthieu.verlynde@univ-smb.fr)
• Ammar Mian (ammar.mian@univ-smb.fr)
• Yajing Yan (yajing.yan@univ-smb.fr)

References

[1] A. Mian, G. Ginolhac, J.-P. Ovarlez, and A. M. Atto, “New Robust Statistics for Change Detection in Time Series of Multivariate SAR Images,” IEEE Transactions on Signal Processing, vol. 67, no. 2, pp. 520–534, Jan. 2019. [Online]. Available: https://ieeexplore.ieee.org/document/8552453/
[2] A. D. C. Nascimento, A. C. Frery, and R. J. Cintra, “Detecting changes in fully polarimetric sar imagery with statistical information theory,” IEEE Transactions on Geoscience and Remote Sensing, vol. 57, no. 3, pp. 1380–1392, 2019.