preparing module for roof top segmentation task
├── README.md <- The top-level README for basic information about using this project.
├── data <- Not tracked by git so add the data to this folder
│ └── raw <- The original, immutable data dump.
│ └── images <- Original images as .png files
│ └── labels <- Original labels as .png files
│ └── test_images <- Test images that don't have labels
│
├── models <- Trained and serialized models, configuration used and model training and validation logs
│
├── notebooks <- Jupyter notebooks. Naming convention is a number (for ordering), and a short `-` delimited description,
│ e.g.`00-data-exploration`.
│
├── reports <- Generated analysis.
│ └── figures <- Generated graphics and figures to be used in reporting
│ └── documentation.md <- Report on approach and performance
│
├── requirements.txt <- The requirements file for reproducing the analysis environment, e.g.
│ generated with `pipreqs roof-top-segmentation`
│
├── environment.yml <- conda exported environment file
│
├── src <- Source code for use in this project.
│ ├── __init__.py <- Makes src a Python module
│ │
│ ├── data <- Scripts to create dataset and dataloaders
│ │
│ ├── models <- Scripts to define model classes and training loop
│ │
│ ├── visualization <- Scripts to create training and results visualizations
│ │
│ ├── main.py <- Main script to start to run train-predict-evaluate pipeline
│ │
│ ├── predict.py <- Predict module used in main.py or can be used independently
│ │
│ ├── utils.py <- Utility functions
│ │
│ ├── config.yaml <- Set configurations for training
Project adapted from the cookiecutter data science project template
To run training pipeline, cd to src and run main.py: "python3 main.py"
After having a saved model in models/experiment_{num}/weights.pt, predict.py in src can be called to generate labels for new images stored in test_images folder under data/raw/ or can also generate labels for train or val images and compare predictions vs the ground truth label.
The visualizations of the loss and metrics are stored in respective experiment folders eg. models/experiment_{num}. The visualizations of predictions can be found in models/experiment_{num}/test_predictions for data/raw/test_images. The train and validation data predictions are stored in models/experiment_{num}/train_predictions and models/experiment_{num}/val_predictions
The project documentation can be found under reports as documentation.md