Data

Tools

Indexes

Applications

Experiments

Open Source Science

satellite-pattern-analysis-for-renewables-spar-tool

Solarfarm and PV segmentation using satellite and aerial imagery

https://github.com/mahdikoubaa1/satellite-pattern-analysis-for-renewables-spar-tool

Science Score: 36.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • ○
    CITATION.cff file
  • ✓
    codemeta.json file
    Found codemeta.json file
  • ✓
    .zenodo.json file
    Found .zenodo.json file
  • ○
    DOI references
  • ✓
    Academic publication links
    Links to: ieee.org
  • ○
    Academic email domains
  • ○
    Institutional organization owner
  • ○
    JOSS paper metadata
  • ○
    Scientific vocabulary similarity
    Low similarity (9.1%) to scientific vocabulary
Last synced: 7 months ago · JSON representation

Repository

Solarfarm and PV segmentation using satellite and aerial imagery

Basic Info
  • Host: GitHub
  • Owner: mahdikoubaa1
  • License: mit
  • Language: Jupyter Notebook
  • Default Branch: main
  • Size: 82.1 MB
Statistics
  • Stars: 0
  • Watchers: 1
  • Forks: 0
  • Open Issues: 0
  • Releases: 0
Created 12 months ago · Last pushed 8 months ago
Metadata Files
Readme License Citation

README.md

Aerial and Sentinel-2 (S2) PV Segmentation

Aerial Segmentation models

Segformer architecture using mit-b5 encoder finetuned on multi-scale PV dataset (link).

SolarSAM architecture using SAM encoder trained on bavarian PV dataset (link).

Model for S2 raster solarfarm segmentation (PIXNN)

Brief description

The model can be seen as a feed forward network having as input the bands of a single pixel bands i.e. we classify by colors/ by band values

In practice

  • we are using ADAM optimizer with a training dataset of n pixels flattened into 1 x n image

  • We are using 1x1 convolution to mimic the behavior of fully connected nn and facilitate the transition to images of arbitrary size.

Training Loss

Focal Loss

Sampling

  • The training dataset is a set of pixels that we sample from labeled polygons drawn on the raster.
  • Pixels sampled from a polygon get the same label (solarfarm or not solarfarm) as the polygon.
  • Each sample contains 13 values.
  • For the best model variant we are using all 13 bands except band 11 to train the model. ## SpaR tool The SpaR tool is a UI interface available in notebooks/SpaR_Tool.ipynb. ### Supported Features
  • Sampling, Training and Prediction for PIXNN architecture (S2 Rasters PV Segmentation).
  • Prediction for SolarSAM and Segformer architectures (Aerial Rasters PV Segmentaion (currently limited to Bavaria)). ## Environment requirement installation (miniconda/anaconda is required) bash bash ./setup_environment.sh ## Data Link For the SpaR tool to work as intended, please extract data.zip in the root directory of this project

Owner

  • Login: mahdikoubaa1
  • Kind: user

GitHub Events

Total
  • Member event: 1
  • Push event: 29
  • Create event: 2
Last Year
  • Member event: 1
  • Push event: 29
  • Create event: 2