2022/07/15 by Gabriel Kasmi, Kasmi, Gabriel, Laurent Dubus +4 · 1 citation
Computer Science · Energy · Environmental Science · #Solar Radiation and Photovoltaics #Photovoltaic System Optimization Techniques #Photovoltaic Systems and Sustainability
paper · pdf · doi:10.48550/arxiv.2207.07466
<strong>Overview</strong> Supplementary material necessary to run the demonstration notebooks from the conference paper "Towards unsupervised assessment with open-source data of the accuracy of deep learning-based distributed PV mapping", presented at the workshop on Machine Learning for Earth Observation, in conjunction with the ECML/PKDD 2022. <strong>Data records</strong> The dataset records are the following : <strong>source/ </strong>root data folder <strong>COMMUNES2021.zip</strong>: the folder containing the shapefile of the French cities, <strong>IGNORTHO202069.zip</strong>: the original folder containing the orthoimagery. This folder was originally downloaded from IGN's Geoportail (see sources below). <strong>IGNTOPO202169.zip</strong>: a folder containing the topological data. This folder is a copy of a folder downloaded from IGN's Geoportail (see sources below). <strong>LOOKUPTABLE.zip</strong>: a folder containing the look-up table, used to infer the tilt angle of the installations based on their location and surface. <strong>RNI2020.zip</strong>: a folder containing the RNI for the year 2020, used as a benchmark to assess the accuracy of the detection. <strong>WEIGHTS.zip</strong>: the weights of our models (segmentation and classification branches). <strong>Usage notes</strong> These source files are intended to be used in the notebook <strong>hands-on.ipynb</strong> accessible here: https://github.com/gabrielkasmi/deeppvmapper/blob/main/notebooks/hands-on.ipynb You can either download the data manually, or directly from the notebook. It also contains the fined-tuned weights for the classification and segmentation models if you want to use them in your own projects. <strong>Sources </strong> The orthoimagery and the topological data are provided by the IGN under an open license. The complete data records can be accessed here for the images and here for the topological data. The RNI can be accessed here and is also provided under open access. The cities' shapefiles also provided under open access, are accessible here.