How to

Data structure

• <repo_home>
  • data
    • images (contains the images)
      • a.png
      • b.png
    • labels (contains the binary segmentation results as bw images)
      • a.png
      • b.png

Train

Run TrainScript.py to train on the dataset in ./data .

• checkpoints will be dumped to ./checkpoints
• tensorboard logs will be dumped to ./logs

Eval

Run EvaluationScript.py to dump segmentation results for all images in the ./data folder. Run "streamlit run ThresholdVisualizationTool.py" to get an interactive gui for result evaluation.