TAMMI_S1S2 — Sentinel-1 and Sentinel-2 Mosaics

This repository contains the department-level Sentinel-1 (SAR) and Sentinel-2 (multispectral) mosaics used to build the TAMMI dataset.

---

Structure

{dept}/
└── S1S2/
    ├── {S2_prefix}TCI.jp2       # Sentinel-2 True Color (RGB, 10m)
    ├── {S2_prefix}B05.jp2       # Sentinel-2 Red Edge 1 (20m)
    ├── {S2_prefix}B06.jp2       # Sentinel-2 Red Edge 2 (20m)
    ├── {S2_prefix}B07.jp2       # Sentinel-2 Red Edge 3 (20m)
    ├── {S2_prefix}B08.jp2       # Sentinel-2 NIR (10m)
    ├── {S2_prefix}B8A.jp2       # Sentinel-2 Narrow NIR (20m)
    ├── {S2_prefix}B11.jp2       # Sentinel-2 SWIR 1 (20m)
    ├── {S2_prefix}B12.jp2       # Sentinel-2 SWIR 2 (20m)
    ├── {S1_vh_name}             # Sentinel-1 VH polarization
    ├── {S1_vv_name}             # Sentinel-1 VV polarization
    └── ratio{S1_vv_name}        # Sentinel-1 VV/VH ratio

Departments available: 75, 92, 93, 94, 74, 34

**Note:** Departments 92, 93, and 94 share the same S1/S2 mosaics as department 75.

---

How to Extract Patches

Use the provided get_s1s2_patches.py script to extract S1/S2 patches for each sample in TAMMI. It uses the s2_center_pos, s1_center_pos, orbit_direction, and look_direction fields from the TAMMI dataset to crop the correct patch from the mosaic.

Installation

pip install datasets rasterio tifffile huggingface_hub scipy numpy

Usage

1. Download the mosaics from this repo:

from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="HichemBoussaid/TAMMI_S1S2",
    repo_type="dataset",
    local_dir="./S1S2_mosaics"
)

2. Edit get_s1s2_patches.py to set:

   • S1S2_LOCAL — path where you downloaded the mosaics
   • OUTPUT_DIR — where patches will be saved
   • SPLIT — "train", "val", or "test"

3. Run:

python get_s1s2_patches.py

The script is resumable — if interrupted, rerunning it will skip already completed patches.

Output

For each image in TAMMI, the script produces:

• S1.tif — SAR patch (H, W, 3) float32 — channels: VV, VH, ratio
• S2.tif — Multispectral patch (H, W, 10) — channels: TCI(RGB), B05, B06, B07, B08, B8A, B11, B12

Saved at {OUTPUT_DIR}/{dept}/{image_id}/S1.tif and S2.tif.

---

Companion Dataset

The main TAMMI dataset (BDO images + Q&A pairs + metadata) is available at:

HichemBoussaid/TAMMI

---

Citation

@inproceedings{boussaid2025tammi,
  title     = {Visual Question Answering on Multiple Remote Sensing Image Modalities},
  author    = {Boussaid, Hichem and Tosato, Lucrezia and Weissgerber, Flora and Kurtz, Camille and Wendling, Laurent and Lobry, Sylvain},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
  year      = {2025}
}