Azure FarmBeats samples for Python

This project contains end to end Python samples and Jupyter notebooks which demonstrate creating farms, boundaries, ingesting satellite and weather data, and building Machine Learning (ML) models related to agricultural applications. It also contains a set of utilities for data processing.

These samples leverage Azure FarmBeats and its corresponding Python SDK and it runs on Azure Machine Learning Compute.

Prerequisites

To run these samples, you must have:

• Azure subscription - Create a free account
• Azure FarmBeats resource - Install FarmBeats
• A subscription with one of the supported weather data providers for FarmBeats - Weather Integration Docs
• Azure Machine Learning (AML) compute resource - Create AML Compute (with Python 3.6)

Features & use case

This project aims to demonstrate following:

Creation of farms, boundaries and ingesting satellite & weather data:

• quick_start.ipynb notebook demonstrates how to create farms, boundaries, and ingestion jobs for satellite and weather data. A satellite ingestion job ingests satellite data (from Sentinel-2) to Azure FarmBeats data store for a given area of interest(AOI) and duration. Similarly, weather job ingests data into Azure Farmbeats store from weather data providers (example: DTN ClearAg, DTN Content Services)

Build NDVI forecasting model:

• An end to end ML model sample that demonstrates how to pull satellite and weather data for any number of farms using Azure FarmBeats SDK, how to create analysis ready datasets (ARD), ML model building, training and deploying ML model to Azure using Azure Machine Learning.

Notebook	Description
1_download_data.ipynb	This notebook demonstrates how to download satellite and weather data for given boundaries to build NDVI forecast model.
2_train.ipynb	This notebook demonstrates building end to end deep learning model using satellite and weather data.
3_test.ipynb	In this notebook, the model forecasts NDVI for next 10 days for an 'Area of Interest' (AOI).
4_deploy_azure.ipynb	This notebook demonstrates how to deploy model and create webservice using Azure ML SDK.
5_inference.ipynb	This notebook demonstrates model inference on a new AOI using the AzureML webservice endpoint and generates NDVI forecast for the next 10 days.

Getting Started

Steps to run samples and build an NDVI forecasting model

1. Launch the terminal of AML compute.
2. Run the following command git clone https://github.com/Azure-Samples/azure-farmBeats-samples.
3. Open Azure Command Line Interface (CLI) and execute following commands to create an environment with all required libraries.
   a. cd azure-farmbeats-samples (Go to azure-farmbeats-samples folder)
   b. conda env create -f environment.yml (Create a new environemt 'farmbeats-sdk-env' with all required packages)
   c. conda activate farmbeats-sdk-env
   d. python -m ipykernel install --user --name=farmbeats-sdk-env (Activate the farmbeats-sdk-env' kernel in AML. Each notebook should select this kernel before running)
4. Go to Notebooks pane and click on folder azure-farmbeats-samples under your alias.
5. Select compute and jupyter kernel ("farmbeats-sdk-env").
6. Update utils/config.py with FarmBeats credentials.
7. Go to folder ndvi_forecast and run the notebooks enumerated in the table above starting with 1_download_data.ipynb.

Contributing

Please refer to CONTRIBUTING.md