Since connectivity from farm to the cloud is limited, it is difficult to send large amounts of data from the farm to the cloud. A single 15-minute drone flight can generate GBs of data, which are difficult to send to the cloud from the middle of a farm.

Our key insight is that most farmers have a PC. If they don’t, we have a PC form factor device that is loaded with a lot of software that does AI, Computer Vision, and processing on an Edge device in the farm.

We have been developing technologies to process large amounts of data in the farm itself – either in the farmer’s house or office, or next to the base stations. The Edge device gets data from sensors, from drones, performs computer vision techniques to stitch drone imagery, and can also run AI at the Edge. We have also developed techniques to intelligently send large amounts of data from the Edge to the Cloud.

Read the paper presented at MobiCom 2021 >

Read the paper presented at NSDI 2017 >

Note: FarmVibes.Edge will be released to GitHub soon.

The post FarmVibes.Edge appeared first on Microsoft Research.

A key challenge for data-driven agriculture is affordable connectivity in the middle of the farm. Even though connectivity might exist in the farmer’s house, there is little to no connectivity in the middle of the field. This is a challenge in many parts of the world.  

FarmVibes.Connect involves inventing new ways to bring Internet to the middle of a farm, including:

• Leveraging unused TV channels for broadband connectivity — This technology refers to unused TV spectrum, in the very high frequency (VHF) and ultrahigh frequency (UHF) bands, which are legal to use in the United States, Canada, and several countries worldwide. Devices using this technology consult with a database to determine the available channels at a location and operate in an available TV channel.
  Read the paper presented at NSDI 2017 >
• Narrowband IoT connectivity in the middle of the farm — LoRa and SIGFOX IoT networks operate around the 2.4-GHz or 900-MHz band, which have limited propagation through crops and canopies. They can operate in 400-MHz “semi licensed band” in the United States, but that needs coordination with a local agency to get a very small sliver—12.5 KHz—of the spectrum, which can support very few sensors. We propose the use of LoRa over TV spectrum, which is abundantly available in rural areas, for IoT communication. It offers very long-range connectivity over tens of miles, even through crops and canopies. We designed and implemented a narrowband IoT radio that can operate in this spectrum. This opens the door for very large-scale network deployments where a single base station can support hundreds of devices across tens of miles. This led to new FCC regulations allowing the use of narrowband devices in TV spectrum.
  Read the paper presented at NSDI 2022 >
• Space communications — Our research in Networking technologies is also looking at new ways to communicate from space to ground, and through Edge computing in Space.
  Read our early work presented at SIGCOMM 2021 >

Note: FarmVibes.Connect will be released to GitHub soon.

The post FarmVibes.Connect appeared first on Microsoft Research.

One of the key challenges, especially for smallholder farmers, is how to communicate with the farmer. This is needed to get additional data, run surveys, and to communicate insights and feedback. We are building tools to automate the entire process.  

The FarmVibes.Bot platform allows organizations to communicate simply and effectively with any user who has a feature phone (by USSD or SMS) or any user who has a smart phone (WhatsApp, Facebook, Telegram, etc.). 

Organizations can disseminate information, collect field data, and connect users to services. The platform also enables the platform owner to connect to external APIs to crowd in services for the user (e.g., financing, insurance, access to markets, or account information). And, built in data visualization enables a detailed understanding of interactions and activities.

This is in use by more than 500K farmers in sub-Saharan Africa.

• Watch the AGRA video > (opens in new tab)
• Read the Africa Outlook story > (opens in new tab)
• Read the Microsoft News Center article > (opens in new tab)
• Read the Microsoft Africa blog > (opens in new tab)

Note: FarmVibes.Bot will be released to GitHub soon.

The post FarmVibes.Bot appeared first on Microsoft Research.

No single data source gives us the complete information about a farm. Sensors capture temporal data for a location, satellites or drones give spatial information at an instant in time, and neither have enough information extract details about properties below the soil’s surface.

Our key hypothesis is that merging a variety of data sources can help us create the ultimate truth about a farm. Through FarmVibes.AI, we have done this for a variety of data sources, including:

• Heat maps — by combining Sensor + Aerial Imagery, e.g., soil moisture, soil temperature, carbon, weed, nutrients, and soil pH heat maps (accepted to
  Read the paper presented at IEEE Micro 2022 >
• Seeing through clouds — by combining Optical + RADAR imagery to see through Clouds, with a technique called SpaceEye
  Read the paper >
• Super-res satellite imagery — by combining low res, cloud-free satellite imagery (from SpaceEye) with higher res cloudy satellite imagery
• Microclimate prediction — by combining on-farm sensor data with weather predictions from weather services
  Read the paper presented at KDD 2021 >

In addition to research, we are making these tools available to the broader community. Scientists, researchers, and partners can build new workflows leveraging these AI models, to estimate farming practices, the amount of emissions, and the carbon sequestered in soil.  

As an agri/food or sustech data scientist or researcher interacting, fusing, deriving insights from various such spatial-temporal datasets is hard. With FarmVibes.AI, one can develop rich insights for agriculture with ease. They can build models that fuse multiple geospatial and spatiotemporal data from the field to obtain insights regarding its carbon footprint (as determined by tilling, fertilization, and cover crops), the nutrition of food that it is growing (e.g., yield and protein content), and the long-term sustainability of the soil and the water (e.g., detect if topsoil erosion is being arrested and water ways are designed to retain rain and flood water). These insights are not otherwise possible (with regard to robustness, precision and/or accuracy) without the fusion techniques in FarmVibes.AI. You can fuse together satellite imagery (RGB, SAR, multispectral, different resolutions), drone imagery (RGB, multispec, hyperspec), weather data (historic, current, and forecasts), sensor timeseries data (soil as well as atmospheric), and more (e.g., information from industrial vehicles such as tractors).

The post FarmVibes.AI appeared first on Microsoft Research.