This repository was developed as part of the Master's Thesis:
"Shoreline Detection Using Deep Learning Techniques on RGB Imagery"
by Josep Oliver Sansó — MSc in Intelligent Systems, Universitat de les Illes Balears (2025).

---

Reproducibility

This repository contains all code and configuration files used in the master's thesis.
The dataset version used can be found at Zenodo.

To reproduce the experiments, please start from the notebooks in /notebooks/ or the scripts in /src/.
All predictions and dataset splits used for evaluation are stored in /predictions/.

---

Directory Structure

/shoreline-extraction
├── /data/
│   ├── /raw/                  # Original (unprocessed) dataset
│   ├── /splits/               # Dataset splits (train / val / test)
│   └── /patchified/           # Dataset with patches and splits (train / val / test)
│
├── /predictions/              # Model predictions and train/val/test splits
│   ├── /station/              # Visual predictions by coastal station
│   └── /splits/               # Text files listing the dataset split
│
├── /src/                      # Source code: data processing, training, evaluation
│   ├── /data_processing/      # Scripts for generating patches, augmentation, etc.
│   ├── /training/             # Model training scripts and configuration
│   ├── /evaluation/           # Evaluation routines and metrics
│   └── /utils/                # General-purpose utilities
│
├── /models/                   # Models and architectures
│   ├── /unet/                 # U-Net implementation
│   ├── /bilstm/               # Bi-LSTM implementation
│   └── /other-models/         # Other models, if used
│
├── /weights/                  # Pre-trained weights and saved models
│   ├── /unet/                 # U-Net weights
│   ├── /bilstm/               # Bi-LSTM weights
│   └── /other-models/         # Weights for other models
│
├── /notebooks/                # Jupyter notebooks for experimentation and analysis
│   ├── /data-exploration/     # Data exploration analysis
│   ├── /model-training/       # Model training notebooks
│   └── /evaluation/           # Model evaluation notebooks
│
├── .env.example               # Example environment variables file
├── README.md                  # Project overview and documentation
└── requirements.txt           # Project dependencies

---

.env configuration

The project includes a .env.example file used to configure MLflow for experiment tracking.
To enable tracking (locally or via a remote server), rename the file to .env and update the variables with your own configuration.