modelling

This repository comprises works undertaken to evaluate, fine-tune and employ AI models used throughout the project

scope

PV detection & segmentation

• performance assessment of YOLOv8-based detection models for solar panels
  • ariel_ultimate_eval_X_25.ipynb
  • finloop_ultimate_eval_X_25.ipynb
  • pool_ultimate_eval_X_25.ipynb
• segmentation of solar panels on images (YOLO-segmentation model and YOLO+SAM composition)
  • ariel_segment_with_sam.ipynb
  • finloop_segment.ipynb
  • finloop_segment_with_sam.ipynb
  • finloop_segment_comparison.ipynb
  • pool_segment_with_sam.ipynb
  • SAM_segment_comparison.ipynb
  • sam_comparison.csv

PV fine-tuning

• synthetic dataset
  • tuning and baseline results
  • tuning script
• Fresno dataset
  • tuning script
  • high-res img evaluation
  • baseline results

tree coverage detection & segmentation

• performance assessment of tree models (Deepforest+SAM, CLIPSeg, SegFormer)
• tree_area_calculation

analysis

correlation.ipynb

acknowledgements

Obliczenia wykonano z wykorzystaniem komputerów Centrum Informatycznego Trójmiejskiej Akademickiej Sieci Komputerowej (Computations were carried out using the computers of Centre of Informatics Tricity Academic Supercomputer & Network).

PV models:

• finloop
• Ariel Drabkin's
• Andrew Gray's

datasets:

• Fresno
• synthetic
• 3City imagery