https://github.com/inseefrlab/satellite-images-inference
https://github.com/inseefrlab/satellite-images-inference
Last synced: 2 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/inseefrlab/satellite-images-inference
- Owner: InseeFrLab
- License: mit
- Created: 2024-02-01T15:49:26.000Z (almost 2 years ago)
- Default Branch: main
- Last Pushed: 2025-09-19T14:05:43.000Z (3 months ago)
- Last Synced: 2025-09-19T16:25:46.978Z (3 months ago)
- Language: Python
- Size: 377 KB
- Stars: 4
- Watchers: 3
- Forks: 0
- Open Issues: 5
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# 🛰️ Satellite Image Segmentation Inference
This repository contains code for performing segmentation inference on satellite images using deep learning models.
## Table of Contents
- [🛰️ Satellite Image Segmentation Inference](#️-satellite-image-segmentation-inference)
- [Table of Contents](#table-of-contents)
- [🚀 Introduction](#-introduction)
- [🛠️ Usage](#️-usage)
- [🔗 Step 1: Register new images](#-step-1-register-new-images)
- [🧠 Step 2: Run inference via API](#-step-2-run-inference-via-api)
- [🌐 API](#-api)
- [🤖 Automation](#-automation)
- [📜 License](#-license)
## 🚀 Introduction
This project is designed to perform segmentation inference on satellite images using deep learning models. The repository includes scripts for making predictions on satellite images and utilities for image format conversions. 📷🛰️
## 🛠️ Usage
To perform inference on a new set of satellite images stored in the S3 Bucket:
📂 Path: `projet-slums-detection/data-raw/PLEIADES///`
### 🔗 Step 1: Register new images
Before running inference, you must register the new images by linking them to their corresponding geometry polygons stored in the partitioned Parquet file:
📂 Path: `projet-slums-detection/data-raw/PLEIADES/filename-to-polygons/`
Run the following command:
```bash
python -m src.build_filename_to_polygons
```
### 🧠 Step 2: Run inference via API
Once registered, you can run inference on these new images using the API:
```bash
python -m src.make_predictions_from_api --dep --year
```
## 🌐 API
All API-related code is in the app/ folder, built using FastAPI ⚡. The key files include:
- main.py 📌: Defines the API endpoints.
- utils.py 🔧: Contains utility functions for API operations.
## 🤖 Automation
The `argo-workflows/` folder contains templates that enable automation and parallelization of inference across multiple departments and years. ⚡🔄
## 📜 License
This project is licensed under the MIT License. 📄✅