How data-driven farming could transform agriculture | Ranveer Chandra | TEDxUniversityofRochester

Summary

The speaker addresses the critical issue of increasing global food production by 70% by 2050 to feed the growing population. They highlight the challenges of limited arable land and receding water levels, proposing data-driven farming and precision agriculture as solutions. By using technologies like TV white spaces, UAVs, and machine learning, the FarmBeats project aims to reduce the cost of data-driven agriculture, making it accessible to smallholder farmers worldwide. The goal is to improve yields, reduce costs, and benefit the environment by enabling precise, site-specific applications of resources like water and pesticides.

Takeaways

• 🌱 The world's food production needs to increase by 70% by 2050 to feed the growing population, highlighting the urgency of the global food problem.
• 📈 Data-driven farming is a promising approach to significantly increase food production by leveraging data to optimize farming practices.
• 🚜 Precision agriculture, which involves site-specific applications of water and pesticides, can improve yield, reduce costs, and benefit the environment.
• 🌡️ Mapping soil moisture and nutrient levels throughout a farm can enable more precise farming techniques and is crucial for data-driven farming.
• 💡 The Farm Beats project at Microsoft aims to reduce the cost of data-driven agriculture solutions by two orders of magnitude, making it more accessible to farmers.
• 📶 Internet connectivity in farms is often expensive and limited, hindering the adoption of data-driven farming technologies.
• 📡 TV white spaces technology can provide cost-effective connectivity in rural areas, enabling the use of data-driven farming solutions.
• 🤖 Using UAVs (drones) and machine learning, it's possible to create detailed maps of farms with fewer sensors, reducing the cost of data collection.
• 🎈 Tethered helium balloons with cameras can be a low-cost alternative to drones for capturing aerial imagery of farms.
• 🖥️ Local processing of data on a farmer's PC can overcome connectivity issues and provide real-time insights without the need for internet.
• 🔢 Farm Beats has demonstrated high accuracy in predicting soil conditions, which can be as reliable as direct sensor measurements for decision-making.

Q & A

What is the projected increase in world food production needed by 2050 to feed the growing population?

The world's food production needs to increase by 70% by 2050 to feed the growing population.

Why is the increase in food production necessary not just to feed but also to nourish the world's population?

The necessity to nourish the world's population makes the problem of increasing food production even more severe due to limited arable land and receding water levels.

What is data-driven farming and how does it relate to solving the world's food problem?

Data-driven farming refers to the ability to map every farm in the world and overlay it with various data, such as soil moisture and nutrient levels. This approach can enable techniques like precision agriculture, potentially increasing food production.

What is precision agriculture and how does it benefit farming practices?

Precision agriculture is the ability to apply water, pesticides, and other inputs uniformly or site-specifically only where it is needed. It improves yield, reduces costs by using fewer resources, and is environmentally friendly by avoiding overuse of pesticides and water.

What is phenotyping in agriculture and how does it differ from genotyping?

Phenotyping in agriculture is the process of understanding why the same seed variety grows differently in different parts of a farm. Unlike genotyping, which focuses on the genetic makeup of plants, phenotyping observes the physical characteristics and growth patterns influenced by environmental factors.

What is the Farm Beats project and what is its main goal?

The Farm Beats project, led by Microsoft, aims to significantly reduce the cost of data-driven agriculture solutions by two orders of magnitude, making it more accessible to farmers and potentially revolutionizing farming practices.

Why has the adoption of precision agriculture technology been slow despite its benefits?

The slow adoption of precision agriculture technology is primarily due to the high cost of existing data-driven agriculture solutions, which can be prohibitive for many farmers.

What is the TV white spaces technology and how does it benefit agriculture?

TV white spaces technology utilizes unused TV spectrum to provide long-range Wi-Fi-like connectivity. In agriculture, it can be used to connect remote farms to the internet, enabling the transmission of data from sensors and other devices.

How does the use of UAVs or drones contribute to creating accurate farm maps?

UAVs or drones equipped with cameras can quickly cover large areas of a farm and take images. Combined with artificial intelligence and machine learning, these images can be used to interpolate data from a few sensors and predict soil conditions across the entire farm.

What is the challenge with using drones for aerial imagery in remote parts of the world, and how is it being addressed?

The cost of drones and regulatory restrictions can limit their use in remote areas. To address this, tethered helium balloons with mounted smartphones are used to capture aerial imagery at a lower cost and without the need for special permissions.

How does the Farm Beats project handle the issue of limited internet connectivity from the farmer's house to the cloud?

The project uses a PC-based system that processes data from sensors and drones locally, reducing the need for high-speed internet. It also compresses and sends only necessary data to the cloud, ensuring that the system can run even with limited connectivity.

What kind of actionable insights can the Farm Beats system provide to farmers?

The Farm Beats system can provide insights such as soil moisture levels, pH levels, and the growth patterns of crops. It can also monitor animal health and movement, and flag areas that require attention, such as water puddles or acidic soil patches.

How accurate are the predictions made by the Farm Beats system compared to actual sensor measurements?

The predictions made by the Farm Beats system are so close to the actual sensor measurements that they are considered actionable by farmers, providing reliable data for making informed decisions.

Outlines

00:00

🌱 Addressing the Global Food Production Challenge

The speaker begins by highlighting the urgent need to increase the world's food production by 70% by 2050 to accommodate the growing population. The challenge is exacerbated by limited arable land and receding water levels, reminiscent of the Green Revolution. The speaker introduces data-driven farming as a promising solution, which involves mapping farms and overlaying them with data like soil moisture and nutrient levels to enable precision agriculture. Precision agriculture is shown to improve yields, reduce costs, and benefit the environment by applying water and pesticides only where needed. The talk also touches on phenotyping, which could lead to the creation of new genotypes. The speaker's background in computer science and personal experiences in India with agriculture set the stage for the Farm Beats project at Microsoft, aimed at significantly reducing the cost of data-driven agriculture solutions.

05:01

📡 Leveraging TV White Spaces for Farm Connectivity

The speaker discusses the high costs associated with existing data-driven agriculture solutions, particularly the internet connectivity required for remote farms. To address this, the speaker introduces the concept of TV white spaces, which are unused TV channels that can transmit Wi-Fi signals over long distances without interference. This technology, once legalized in 2010, has been used to connect rural areas to the internet. The speaker's insight is that farms, often located away from cities, have many empty TV channels, providing ample unused spectrum for connectivity. The vision is to use TV white spaces to connect entire farms, enabling the collection of data that was previously unattainable. The speaker also explains how this technology can be used to connect various devices on a farm, facilitating the gathering of extensive agricultural data.

10:02

🚁 Utilizing UAVs and Balloons for Aerial Farm Imaging

This section delves into the challenges of creating detailed maps of farms to understand soil conditions and other variables. The speaker proposes the use of Unmanned Aerial Vehicles (UAVs) or drones to quickly cover large areas and take images, which can then be used with artificial intelligence and machine learning to interpolate data from a few sensors across the entire farm. The speaker also addresses the challenges of using drones in certain countries and introduces a low-cost alternative: tethered helium balloons equipped with smartphones to capture continuous imagery. This approach allows for the creation of detailed maps with fewer sensors and provides a solution for areas where drones are restricted or too costly. The speaker describes how this technology has been used to monitor crops and prevent waste due to flood damage, with the potential for adaptation in various parts of the world.

15:03

🖥 On-Farm Processing and Machine Learning for Precision Farming

The speaker addresses the issue of limited internet connectivity from farms to the cloud, which hinders the transfer of large amounts of data collected by drones and sensors. To overcome this, the Farm Beats project utilizes on-farm processing, where a PC or a provided box runs analyses using data from sensors and drones, applying machine learning to generate actionable insights without relying on cloud connectivity. The speaker emphasizes that this approach allows for the retention of much data on the farm itself, reducing the need to send large volumes of data to the cloud. The system can also operate offline, ensuring continuity even during internet outages. The speaker shares examples of how this technology has been deployed and used by farmers to gain quick and valuable insights into their operations, such as monitoring cow health and identifying areas of the farm that require attention.

📊 Demonstrating the Accuracy and Impact of Farm Beats

In the final paragraph, the speaker presents the results of accuracy tests for the Farm Beats system, comparing its predictions for soil temperature, pH, and moisture with actual sensor measurements. The findings show that the predictions are highly accurate and actionable for farmers. The speaker also shares specific examples of how the system has been used to provide real-time insights, such as identifying areas of a farm that need attention before the next planting season and monitoring cow health. The speaker concludes by expressing gratitude for the opportunity to address the global food problem and showcasing the potential of technology to assist in solving it, highlighting the collaboration with farmers and agronomists to develop and implement these solutions.

Keywords

💡Food Production

Food production refers to the process of growing, harvesting, and preparing food. In the context of the video, it is emphasized that food production needs to increase by 70% by 2050 to meet the demands of the growing global population. This is a central theme as it sets the stage for discussing the challenges and solutions in agriculture.

💡Arable Land

Arable land is land that is suitable for agriculture and can be used to grow crops. The script mentions that the amount of arable land is limited, which is a significant challenge to increasing food production. This concept is critical as it highlights the scarcity of resources needed for farming.

💡Data-Driven Farming

Data-driven farming is an approach that utilizes data and analytics to optimize agricultural practices. The video discusses how mapping every farm with detailed data can enable precision agriculture and other techniques. This concept is key to understanding the innovative solutions being proposed to improve food production.

💡Precision Agriculture

Precision agriculture is a farming management technique that uses site-specific applications of inputs like water and pesticides only where they are needed. The video explains that this technique can improve yield, reduce costs, and benefit the environment. It's a core part of the proposed solution to the food problem.

💡Phenotyping

Phenotyping in agriculture is the process of measuring the observable characteristics of crops, such as growth patterns, which can be influenced by the environment. The video suggests that understanding why the same seed variety grows differently in different parts of the farm can lead to the creation of new genotypes. This is an important concept for improving crop yields.

💡TV White Spaces

TV white spaces refer to the unused frequency bands in the television broadcast spectrum. The video describes how these frequencies can be utilized to provide internet connectivity to remote and rural areas, including farms. This technology is highlighted as a solution to the challenge of internet connectivity for data-driven agriculture.

💡UAVs (Drones)

UAVs, or drones, are unmanned aerial vehicles that can be equipped with cameras and sensors to capture imagery and data from large areas quickly. In the video, drones are used to gather aerial imagery for creating detailed maps of farms, which is an essential part of the data-driven farming approach.

💡Artificial Intelligence (AI)

Artificial intelligence refers to the ability of machines to perform tasks that would typically require human intelligence, such as learning and problem-solving. The video discusses using AI and machine learning to interpolate data from a few sensors to predict values across the entire farm. AI is central to the innovative techniques being developed to enhance farming practices.

💡Machine Learning

Machine learning is a subset of AI that involves the use of algorithms to parse data, learn from it, and make informed decisions based on what they've learned. The video mentions using machine learning to create models that predict soil conditions across a farm, which is crucial for precision agriculture.

💡Soil Moisture

Soil moisture is the amount of water contained in the soil. Accurate measurement of soil moisture is important for farming as it affects plant growth and irrigation needs. The video uses soil moisture as an example of the kind of data that can be collected and analyzed through data-driven farming techniques.

💡Cost Reduction

Cost reduction is the process of decreasing expenses while maintaining or improving the quality of output. The video emphasizes the goal of the Farm Beats project to reduce the cost of data-driven agriculture solutions by two orders of magnitude, making it more accessible to farmers, particularly smallholder farmers.

Highlights

World's food production needs to increase by 70% by 2050 to feed the growing population.

Agricultural scientists are focusing on data-driven farming to address the food production challenge.

Data-driven farming involves mapping farms and overlaying them with data like soil moisture and nutrient levels.

Precision agriculture allows for site-specific applications of water and pesticides, improving yield and reducing costs.

Phenotyping can help understand why the same seed variety grows differently in different farm areas.

The Farm Beats project at Microsoft aims to reduce the cost of data-driven agriculture solutions by two orders of magnitude.

TV white spaces technology can provide long-range connectivity for farms, overcoming the cost barrier.

AI and machine learning can predict soil conditions across a farm using aerial imagery and sparse sensor data.

Tethered helium balloons are a low-cost alternative to drones for capturing continuous aerial imagery.

Farm Beats combines TV white spaces, sensors, and drones to create actionable insights for farmers.

Farm Beats technology can run offline, providing resilience against internet connectivity issues.

Farm Beats has been successfully deployed on farms, providing real-time insights on soil conditions and livestock.

The system can flag areas of a farm that need attention, such as water puddles or areas of differing soil pH.

Farm Beats' predictions are highly accurate and actionable, closely matching actual sensor measurements.

Farm Beats can monitor livestock health and movement, providing immediate feedback to farmers.

The project aims to bring these technologies to smallholder farmers worldwide, enhancing global food security.

Farm Beats demonstrates how non-agricultural technologies can be applied to solve pressing agricultural challenges.

The project combines advances in computer science with practical agricultural needs to create innovative solutions.

Transcripts

you

good morning everyone happier today

today I'm here to tell you about the

world's food

problem the world's food production

needs to increase by 70% by 2050 to feed

the growing population of the world and

this is just to feed if you talk about

nourishing the world the problem is even

more severe and the reason this is such

a big problem is because the amount of

arable land is limited the water levels

are receding so it's like the Green

Revolution problem all over again how do

we get to this significant increase in

food production the agricultural

scientists have been think about this

problem for quite a while and the most

promising approach right now seems to be

that of data-driven farming what we mean

by data-driven farming is the ability to

map every farm in the world and overlay

it with lots and lots of data for

example what is my soil moisture level

six inches below the soil throughout the

farm what is my soil nutrient level

throughout the farm if you could build

maps like this this could enable

techniques like precision agriculture

what we mean by precision agriculture is

the ability to do site-specific

applications for example right now

farmers they apply water uniformly

throughout the farm they'd apply

pesticide uniformly throughout the farm

with precision agriculture you could

apply it only where it is needed

precision agriculture as a technique has

been shown to improve yield reduce cost

because farmers would use less water

less pesticide it's also better for the

environment because you're not putting

in more pesticide than needed you're not

putting in more nitrogen you're not

wasting water the other technique that

such maps could enable is called

phenotyping just like you could do

genotyping you could do phenotyping as

well that is if you could understand why

did the same seed variety grow

differently in different parts of the

farm for example in the red or blue

parts of the farm you could then create

new genotypes this map that you're

seeing here is one such map that we want

to create for all farms in the world so

in the rest of this talk I will talk

about precision agriculture but you can

see how the same techniques apply for

phenotyping as well so precision

agriculture as a technique given that

the benefits are known was first

proposed back in the 80s it's been 30

years since then

the technology hasn't taken off the

biggest reason this technology hasn't

taken off is because of the cost of

existing data-driven agriculture

solutions just to give you an idea of

how expensive it is I was at an expo at

a university where there were several

companies talking about the latest

precision AG equipment the latest sensor

equipment the cheapest sensors that were

available there were five sensors for

$8,000 and a recurring cost for a farmer

to afford that kind of equipment is

expecting too much especially when they

don't know what is the other wife what

is my return of investment if I buy

these sensors which is so expensive that

is the goal of the farm beets project

that I'm leading at Microsoft our goal

is to bring down the cost of these data

driven agriculture solutions by two

orders of magnitude we want to bring it

down from eight thousand to 80 and I'll

talk about a few techniques that we

think we can get that I think we can

help us get us get there before I go

that I wanted to let you know that I

don't have a farming background my

background is a PhD in computer science

but the first 18 years growing up I grew

up in India and as it happens in India

people from India can relate to it we

used to spend we had three brothers and

a sister and we used to spend time with

her grandparents in northern part of

India in a small village in Bihar and we

used to go there spend time in farms by

the way I did not like farming the fact

that those were the worst four months of

my life every year but the reason but

what the reason was that there was no

electricity no toilets it was like in

going from a city to spend four months

in a village wasn't exciting but in a

way that kind of exposed me to the

problems of Agriculture and that's one

of the goals of the farm beets project

as well we want to take the technologies

that we are building to the smallholder

farmers everywhere in the world so I'll

start by talking about the US but you

can see how this relates to other parts

of the world as well so going back the

goal of the farm beets project is to

bring down the costs of data-driven

agriculture solutions

significantly from where they are and

I'll talk about three challenges because

of which existing solutions are

expensive the first reason existing

solutions are expensive is because of

Internet connectivity the farmers house

in this case has some sort of

connectivity to the Internet they pay

for broadband they

one two three megabits a second but the

actual farm is a few miles away

the reason existing solutions are

expensive is because the farms don't

have connectivity they end up using

satellite or custom cellular solutions

to connect these devices to the Internet

so how do we bring down the cost of

connectivity from the middle of the farm

to do this we use one of the prior

research I started researching on this

concept in 2005 called the TV white

spaces what the TV white spaces enables

is imagine if you go by a Wi-Fi router

and plug it in your house imagine if you

could access it a few miles away that

would be cool right as soon as you exit

your house the Wi-Fi connection just

disappears the way we do that is we took

a Wi-Fi signal and put it in empty TV

channels

this is over-the-air TV so you know when

you're a browsing through over-the-air

TV on certain channels you see some

reception other channels all you see is

white noise there's nothing coming there

with this technology we were able to fit

a Wi-Fi signal in those empty TV

channels and noisy TV channels in a way

that did not interfere with the

reception in an adjacent channel so you

could be watching channel 7 at home on

channel 8 we could be sending Wi-Fi

signals and the reason this is so cool

is that compared to Wi-Fi at the same

power level in UHF TV frequencies your

signals go four times father-in VHF they

go 12 times father and this is just

based on pure physics once you put trees

crops canopies and so on your signals

just keep going through in our latest

experiments we put these sensors in soil

of a meter under soil and a signal just

keeps going through so this was a

technology we had built back into 2009

2010 is when the FCC Chairman had come

to visit Microsoft to see the demo we

had put together this was made legal in

the US in 2010

since then we have been deploying this

technology in several parts of the world

connecting rural hospitals schools

libraries to the internet using this

technology

just to recap this is what the TV white

spaces is in Seattle each of these holes

there these gaps are what is the empty

TV spectrum and I already told you how

it is better than Wi-Fi or any

the technology that exists in the

context of agriculture our key insight

was that TV towers are where people are

in Rochester you'll have TV towers in

New York City you'll have TV towers the

farms are away from the cities there are

fewer people so if you go to a farm and

turn on an over-the-air TV you'll find

very few channels most of the channels

are just white noise the more such empty

channels you have the more capacity you

have so if you go to a farm you have a

lot of unused spectrum we are talking of

hundreds of megabits per second of

unused capacity at which point we are

not only talking of connecting sensors

you could be connecting cameras drones

tractors you could be getting a lot of

information that you previously couldn't

get if you talk to any agricultural

scientists the number one problem

they'll talk about is data how do you

get data from the middle of the farm

with the TV white spaces we believe we

can solve that problem our vision here

is just like Wi-Fi connects your house

this TV white spaces could be used to

connect your entire farm in fact this is

what we are doing right now in our

deployments we put this antenna in and

miles around it now gets connected

you're able to get data that you

previously just couldn't gather so this

was challenge number one

the second challenge as I'd mentioned

what we want to get to are these kind of

maps what is the soil moisture level six

inches below the soil throughout the

farm how do we get there if you wanted

to build an accurate map like this you

need lots and lots of sensors you'll

probably need a sensor every 10 meters

but putting a sense that every 10 meters

is expensive to deploy to manage it'll

come in the way of the farmer as the

farmer does to day to day job so the key

question is the last bullet over here on

this slide can we build such a map using

very few sensors the way we solve this

problem is using UAVs these are drones

which can fly large areas very quickly

they have a camera at the bottom that

that can take images of the entire farm

the key technology we built using

artificial intelligence and machine

learning techniques was a way to use the

aerial image to interpolate the data

from a few sensors and predict what

these values are in other parts of the

farm just to give you an idea the state

of the art if people had to build maps

like maps such as the one I showed in

the previous

one would need one would put a few

sensors and then use either linear

interpolation or Cragen's method is a

state of the art but just use that to do

the prediction with this technology you

can use the Adel image the key insight

is that if two parts of the farm looks

similar either in RGB hyperspectral

multispectral imagery they're likely to

have similar values so this was one of

the key algorithms that we developed one

we showed this to work very well for

soil moisture soil temperature and pH

once we started talking about it there

were various companies that came to us

and told us to apply this in other for

other services as well one of the key

challenges in aerial imagery using UAVs

well we use that in the US but as I said

we want to take it to the remotest parts

of the world

well UAVs are great these drones the

ones that you can buy for $1,000 but

thousand dollars is still a lot of money

if you think of smallholder farmers in

Africa in India other problems with

drones are for example in some of the

countries where we want to get aerial

imagery if we wanted to you if we wanted

to use a drone we needed to get

permission from the Ministry of Defense

well at that point it isn't happening so

then how do we get aerial imagery at

low-cost

the way we do that is we have a low-tech

solution we use tethered helium balloons

which are tethered to the ground they

fly up 250 and 200 feet and they are

able to take continuous Edel imagery of

the farm they can last up to four to

seven days the particular thing we built

was a custom mount on which you can put

a smartphone and a battery pack and this

thing can keep clicking pictures for a

really long period of time there is a

farmer in Washington 25 miles east of

Microsoft campus with whom we work he

uses this technology to monitor floods

so one of the problems either he's a

smallholder farmer he sells his produce

to the farmers market and to restaurants

one of the problems he runs into is

every time there is a flood he needs to

throw away all his crop because

regulations require that any crop that

is touched by the flood needs to be

thrown away right now that's what he

does every morning when he comes he sees

there was a flood he throws away all his

crop with this technology he can monitor

he knows which crops had actually

touched by the flood and only throws

away those crops in places like India

and Africa someone could just walk

around with the balloon or put it on a

bike or a tractor and

we have computer vision algorithms based

on which we can stitch this together to

create these big aerial maps for the

entire park the key challenge here was

that with drones you can keep them

stable with balloons they'll move around

with Reindeer camera is not always

facing down so we have computer software

based on which we are able to solve

these problems so going back how do we

build those accurate maps you'll get the

aerial image either from drones or from

balloons we then build these beautiful

pictures of the entire farm we then take

the raw sensor data and build the

machine learning algorithm the

artificial intelligence of them that's a

model of how would these variables

propagate throughout the farm and then

use that to predict what these values

are throughout the entire farm as I

mentioned we've done this for pH

moisture and temperature and we're

working on other variables as well the

key takeaway here is that right now if

you look at the startups in the

agriculture AG tech space a lot of them

are working on either sensors or drones

we believe you are the first who's been

able to combine them in a meaningful way

but we're just starting on the space

there's a lot more to be done but this

seems like a very promising approach to

build these maps at low costs with the

TV white spaces you can bring down the

cost of each sensor with this technology

you need much fewer sensors than what

you would otherwise need to build

accurate maps like this the third

challenge is I mentioned how you can

gather a lot of data and bring it to the

farmers house over the TV white spaces

apply machine learning but the

connectivity from the farmers house to

the cloud is not that great many farmers

they pay for broadband but all they get

is one to three megabits per second

connectivity in your house just to give

an idea of how limiting that is if you

fly a UAV a drone in 15 minutes you

could be generating over a gigabyte of

data you can't send that to the cloud

over a 1 to 3 megabits per second

connection it will take a long time the

other challenge is this connectivity is

also prone to outages so there's a

farmer in eastern New York on the

Vermont border at whose farm we've

deployed a system every time there is a

snowstorm there's a high likelihood that

his internet connection goes off so in

that case even if we connect the data

but we can't provide this information to

the farmer it's of no use

so to solve this problem what we do our

key insight was that most farmers have

pcs if they don't have a PC we ship them

a box

when everything inside this blue box

runs on the farmers PC takes data from

sensors from drones it's able to over

the TV white spaces it then does a lot

of this aerial image regeneration it

does the machine learning piece the heat

map generation all of that in the

farmer's house itself it then we then

feed this into the AG services that the

blocks and blue are the ones we've

already built the other ones are the

ones we were working on of course as I

mentioned my background is not in

agriculture so we partner with people

who understand agriculture the

agronomist farmers to build all those

services the other thing is that a lot

of data stays in the farmers house we

are not sending all the data to the

cloud for example the detailed drawn

imagery the gigabytes of video you can't

ship all of that to the cloud but we

transport it compress it and send a

compressed version to the cloud so these

are some of the unique features on the

left of this entire system it can also

run offline we can disconnect the system

from the year and their Ethernet and it

continues to run so just to give you an

idea of what or how farmers are using it

we've deployed it in quite a few farms

now I'll talk about two of them one is a

small farm it close to Microsoft campus

the small older farmer and the other is

a 2,000 acre farm in upstate New York

the kind of insights we can provide to

the farmers or I'll walk you through

some of them this is a full kilometer

stretch the farmer in New York he wanted

to know how his cows are doing we flew

the drone and within 30 minutes we

transferred the data with the TV white

spaces to this PC and we can start

generating insights like the grass is

growing back well from left to right

there is a water puddle that needs to be

fixed before the next planting season

the cows are pooping well which is also

important information for the farmer

this is where the cows are and this is a

stray cow that needs to be herded in all

of this within 30 minutes of flying the

drone the state-of-the-art is people

would fly the drone take the SD card out

go to a city upload all the data wait

for 24 hours by that time the strake I

would have gone somewhere else with this

we can generate all of this within 30

minutes other kind of insights this is

the farm incarnation we are able to show

the farmer beautiful pictures like this

this is a soil moisture map where we

were able to flag that the top left

corner of the farm is still moist even

though we did not have a

said over there this is after the farmer

had applied lime we were able to flag

that the dark parts that you see here

are still acidic the key question you

would ask is how accurate is this so we

went to three farms three different five

acre plots and we captured a thousand

measurements and we asked the question

that if we pick just ten of these how

accurately can we predict their 990 are

the values using the farm beats

technique of using aerial imagery with

drone sensor data so these are the

results for temperature pH and moisture

this is how accurate farm beats is and

this is how accurate the actual sensors

were that is the sensors for example we

use for soil temperature would report

temperature in one degree Fahrenheit and

so on so the key takeaway here is not

that we are more or less accurate than

the actual sensors themselves but that

are our predictions are so close to the

actual measurements that they are

actually actionable by the farmer then

again another scenario was we had

cameras in barns streaming data with the

TV white spaces very able to flag cows

or the cows are moving around well

whether some cow is sick all of this

because we are doing the TV white spaces

and we have a PC where we are running

all the all the analysis in the farm

itself so you can that's what farm beats

is I wanted to make you aware of the

food problem and some of the steps that

we are taking even not being in the

agriculture space of how we can help

solve some of the food problem with the

TV white spaces we are able to gather a

lot of data that didn't exist and

applying the latest advances in

artificial intelligence and machine

learning we can bring actionable

insights to the farmers so these are the

two farmers we work very closely with

the three farmers Shawn mark and Kristen

just amazing people so with that I

wanted to end the talk thank you

[Applause]