v0.1.1

• Detector
  • MlPreprocessing
  • Docker image lightweight package building using github actions
• Experiments
  • Run tests with results f1score: 0.8382 - Non ML, 0.8644 - ML
• Neural Networks
  • Learned model for PV modules semantic segmentation to 96.17 f1score.

0.1.0

• Detector
  • Extracting thermal image data from raw images
  • Simple CLI interface
  • Using classical image processing techniques detecting PV panels
  • Predefined 2 configuration files - PlasmaConfig, JetConfig
  • Predefined 4 labellers - TiffLabeler, YoloLabeler, LabelMeLabeler, PickleLabeler
  • Docker image lightweight package
• Experiments
  • Created automatic evaluation tool with fscore calculation
  • Labeled PV Modules dataset publicly available
• Neural Networks
  • Learning Pipeline for 4 neural networks architectures based on UNET
  • Labeled PV Panels dataset publicly available