The MaLiSat Toolbox (Beta) 🌊🛰️

Authors: I. Pereira-Sánchez, D. Torres, F. Alcover, B. Garau, C. Alomar, B. Coll, J. Navarro, C. Sbert, S. Deudero, J. Duran.

Sea2Net is a research tool designed to streamline the workflow of working with Sentinel-2 (S2) satellite imagery for marine applications.
Given a set of coordinates and a time period, the app:

1. Fetches Sentinel-2 products available over the region and time span.

2. Upsamples the 20 m bands to 10 m resolution using a guided-image super-resolution network (GINet).

   The GINet is the architecture proposed in Super-Resolution of Sentinel-2 Images Using a Geometry-Guided Back-Projection Network with Self-Attention, available on arXiv:

   

3. Feeds the enhanced images into a segmentation network to detect marine litter.

   The TAUNet is the architecture proposed in Transformer Assisted U-Net for Marine Litter Detection on Sentinel-2 Imagery, available on EarthArXiv:

   

Features

• 📍 Input geographical coordinates
• ⏳ Choose a time interval
• ⬇️ Automatic download of Sentinel-2 products
• 🔍 Super-resolution of 20 m bands with GINet
• 🧪 Experimental stage (alpha)

---

User Guide

Follow these steps to get started with the MaLiSat Toolbox

1. Clone the repository

git clone https://github.com/TAMI-UIB/S2API.git
cd S2API

2. Create a virtual environment (optional but recommended)

# Create a virtual environment
python -m venv venv

# Activate it
# Windows
venv\Scripts\activate
# Linux / Mac
source venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

⚠️ Make sure your Python version is compatible (Python 3.10 or higher recommended).

4. Run the Launcher

python launcher.py

• The launcher will open a GUI asking whether to download new Sentinel-2 products or fuse existing products (in the following version, a marine litter detection option will be added).
• For downloading, you’ll enter coordinates, select a date range, set max cloud cover, and choose a save directory.
• If products are found, the app will ask how many to download.
• For fusing, you just select the folder containing previously downloaded products.

5. Notes

• Internet connection required for downloading Sentinel-2 imagery.
• The checkpoints folder must be present for the fuser (checkpoints/GINet_best.ckpt). The default path is automatically detected relative to the repo.
• Output files are saved in the chosen folder (default: ~/BandesAPP/).
• Running on GPU is strongly recommended.

---

Roadmap / To-Dos

• Integrate segmentation network for marine litter detection
• Improve data handling for large areas and long time spans
• Build a user-friendly interface (CLI / web app)
• Add unit tests and benchmarking against baseline methods

Acknowledgements

This work was funded by MCIN/AEI/10.13039/501100011033/ and by the European Union NextGenerationEU/PRTR via the MaLiSat project TED2021-132644B-I00.