Themed Section : Engineering and Technology

© 2018 IJSRSET | Volume 5 | Issue 3 | Print ISSN: 2395-1990 | Online ISSN: 2394-4099

A Study on Deep Learning in Agriculture

Abhishek Pandey1, Dr. V. Ramesh2
1Research Scholar, SCSVMV University, Kanchipuram, Tamil Nadu, India
2 Assistant Professor, SCSVMV University, Kanchipuram, Tamil Nadu, India
ABSTRACT

We have seen in recent years an amazing improvement in applications using Deep learning. It started with
speech recognition then moved on to computer vision, object recognition and natural language processing.
Deep learning constitutes a recent, modern technique for image processing and data analysis, with promising
results and large potential. Deep learning are machine learning algorithms based on learning multiple level of
abstraction. As deep learning has been successfully applied in various domains, it has recently entered also the
domain of agriculture. In this paper, we explored the platforms that employ deep learning techniques, applied
to various agricultural and food production challenges. We examine the particular agricultural problems under
study, the specific models and frameworks employed the sources, nature and pre-processing of data used, and
the overall performance achieved according to the metrics used at each work under study. Moreover, we study
comparisons of deep learning with other existing popular techniques, in respect to differences in classification
or regression performance. Our research findings indicate that deep learning provides high accuracy,
outperforming existing commonly used image processing techniques.
Keywords: Deep Learning, Classification, Regression

I. INTRODUCTION learning in many fields of life like in Agriculture that

it can monitors crop conditions, like water scarcity
It is expected that global population will reached 9 conditions, plant population in field and soil moisture
billion in 2050.it means we require more agricultural content etc. Deep learning is working in almost all
production in order to meet food demands. Otherwise disciplines of agriculture. In Irrigation it can control
we will be suffering from food security problems. So irrigation water in the field we can optimize the use
we have to increase our per unit area production. In of water because it have automatic irrigation systems
agriculture we emphasis on yield. For increasing yield, in it that take weather conditions and predicts
our farm enterprises require new and innovative amount of water to be applied. Non chemical weed
technologies to face and overcome these difficult control is used in discriminating between weeds and
challenges. So we are using different methodologies, crop seedling. It is an important step towards control
technologies, and different processes for higher yield of weeds by nonchemical way. Drone technology is
per unit area. In methodological approaches we can another adoption of Artificial intelligence. Drones can
use different simulation models, but nowadays deep be used to provide detailed mapping of crops in the
neural network or deep learning is also used in fields. They are also capable of delivering customized
modern countries. Deep learning is a discipline of fertilizers pesticides, insecticides based upon the
Computer Science that deals with giving ability to the requirement of each crops.
machines that it seems to behave like, that it has Artificial intelligence and deep learning algorithms
human intelligence. There is a lot of scope of deep has been found useful in almost every field of work

IJSRSET185318 | Published : 28 April 2018 | March-April-2018 [(5) 3 : 125-128]

125
 and study. It being deployed in every field makes it In this paper we explore applications of deep
the next big thing and breakthrough for a smart learning to provide business leaders with an
future. In India, especially in Punjab increasing understanding of current and emerging trends, and
agriculture output to meet ever increasing present representative examples of popular
population’s demand is one of the major issues being applications. The most popular applications of Deep
faced today. Agricultural experts would tell that Learning in Indian agriculture appear to fall into
factors like climate, soil, rain along with other factors three major categories:
affect the output of a crop. Farmers feel immense
pressure in such situations and lack of knowledge to i. Crop and Soil Monitoring – Companies are
counter the problems faced and modern techniques of leveraging sensors and various IoT-based
farming only add further to the problem. There have technologies to monitor crop and soil health.
been various sorts of efforts to implement deep ii. Predictive Agricultural Analytics – Various AI
learning and check the innumerable effect it has on and machine learning tools are being used to
different aspects like production time, assistance, predict the optimal time to sow seeds, get alerts
output etc. We have seen field of medical science on risks from pest attacks, and more.
specifically benefitting from the implementation of iii. Supply Chain Efficiencies– Companies are using
deep neural networks. Researchers are constantly real-time data analytics on data-streams coming
working of technologies like machine learning, deep from multiple sources to build an efficient and
neural network to aid artificial intelligence as a smart supply chain.
product for human use.
Here we explore various Deep and machine learning
II. RELATED WORK frameworks that helps our farmers to strengthen their
crop production.
If we look specifically towards the field of farming,
quite a lot of research is being conducted. Research A. Intello Labs – Using Deep Learning for Image
on topics like hydration characteristics of Wheat, Analysis
applying machine learning to agricultural data show
that farming activities can greatly benefit from deep Bengaluru-based Intello Labs was started by IIT-
learning. Every research being done aims to improve Bombay alumnus Milan Sharma in May 2016. The
the control of input variables such as fertilizer, seed, company claims to provide advanced image
chemicals or water with respect to the desired recognition technology that can recognize objects,
outcomes of increased profitability, reduced faces, flora and fauna and tag them in any image. The
environmental risk or better product quality. company claim to use deep learning algorithms on
Research on some specific important crops like wheat which a new generation of intelligent applications are
are being carried out to because crops like wheat, rice being built for applications including agriculture, e-
need high output to meet the mass demand and any commerce, advertising, manufacturing, and curation.
case of crop failure means huge loss. So deep learning Small farmers around the world follow traditional
can reduce chances of crop failure due to improper farming practices due to lack of access to scientific
farming activities. This all data is learned and used by understanding of crop lifecycle, pests, quality metrics
deep learning. Hence, a lot of research is being and the latest micro-fertilizers. Intello Labs Image
conducted for the same. based solutions provide insights on the crops’ health
during the growing season. Intello Labs claims to
III. DEEP AGRICULTURE FRAMEWORK provide Agricultural Product Grading: Automated
quality analysis of images of food products is an
International Journal of Scientific Research in Science, Engineering and Technology ( www.ijsrset.com)
126
 accurate and reliable method for grading fresh MAI is the standardized measure used for assessing
products (fruits, grains, vegetables, cotton etc.) the degree of adequacy of rainfall and soil moisture to
characterized by color, size and shape. Their solution meet the potential water requirement of crops.
reads the image that a farmer has taken on his phone Microsoft has also partnered with United
and determines the product quality in real time, Phosphorous (UPL), India’s largest producer of
without any manual intervention. Alerts on Crop agrochemicals, to create the Pest Risk Prediction App
Infestation: Farmers can click an image of their crop that again leverages AI and deep learning to indicate
and use their solution to understand the pests, in advance the risk of pest attack. Today, these
diseases, and foreign plants (weeds) growing in their farmers across the Indian states of Andhra Pradesh
farms. The solution uses deep learning and image and Karnataka wait to get a text message before
processing models to identify any crop diseases or pest sowing the seeds. As per the report cited above, in a
infestation in the crops. Along with the parameters, it few dozen villages in Telangana, Maharashtra, and
gives recommendations on how that disease can be Madhya Pradesh, farmers receive automated voice
cured and prevented from increasing further. calls alerting them whether their crops are at risk of a
pest attack based on weather conditions and stage of
B. Microsoft India – AI-Based Sowing App the crop. No specific numbers on the results were
reported.
Determining the right time to sow crops is often
one of the biggest challenges for Indian farmers C. Gramophone (Agstack Technologies) – Image
where drought and excess rainfall can be equally Recognition for Soil Science
serious challenges. Microsoft in collaboration with
ICRISAT (International Crops Research Institute for Based in the Indian state of Madhya Pradesh,
the Semi-Arid Tropics), developed an AI Sowing App Gramophone claims to leverage the power of image
that uses machine learning and business intelligence recognition and soil science to help farmers with
from the Microsoft Cortana Intelligence Suite. The timely information, technology, and right kind of
app sends sowing advisories to participating farmers inputs to achieve better yields. This technology
on the optimal date to sow. The best part – the platform uses deep learning to predict pest and disease,
farmers don’t need to install any sensors in their fields forecast commodity prices for better price realizations
or incur any capital expenditure. All they need is a and recommends products to farmers. Gramophone
feature phone capable of receiving text messages. use temperature, humidity and pathology/entomology
data to give accurate advisory to the farmers.
Gramophone primary focus is to digitize agriculture
science and convert it into actionable information for
the farmers to provide personalized farm management
solution, which would guide farmers across the
cropping cycle.
Figure 1. A Photo of A Text Message From The
Microsoft India AI Sowing IV. CONCLUSION

To calculate the crop-sowing period, historic climate Given the huge potential of agriculture in India, it’s
data (spanning over 30 years from 1986 to 2015) for imperative that deep and intelligent technology is
the specific area in Andhra Pradesh was analysed used to the maximum so that both farmers and
using AI. To determine the optimal sowing period, consumers can make the most of it. With recent
the Moisture Adequacy Index (MAI) was calculated. advancements in technology coupled with conducive
International Journal of Scientific Research in Science, Engineering and Technology ( www.ijsrset.com)
127
 government policies, we have seen many agtech
startups emerge in the country which is a great
starting point for the penetration of advanced
technologies like AI in agriculture. AI and deep
learning comes as a great boon to the agricultural
sector which is heavily dependent on climatic
conditions which are often unpredictable.
Throughout the paper, we have seen instances of how
companies are using sensors for crop and soil
monitoring, as well as AI and deep learning tools for
predictive analytics. More and more use cases of deep
learning in agriculture are likely to show up in the
near future because of the immense value it can add.

V. REFERENCES

[1] McKinion, J. M., and H. E. Lemmon. "Expert

systems for agriculture." Computers and
Electronics in Agriculture 1.1 (1985): 31-40.
[2] Jat, Avnish Singh, Suma Dawn, Abhinav Mishra,
and Amit Kumar. "Asperger’s Disorder:
Application for its Treatment through Emotion
Detection."
[3] D. Toppeta, ‘‘The Smart City Vision: How
Innovation and ICT can build Smart, Livable,
Sustainable Cities’’, iThink, Report 005, 2010.
[4] Verma Kunal, Pabbi Dinesh, ‘‘Agriculture
Advancement Using Artificial Intelligence’’,
International conference on recent innovations in
science, technology, management and
enironment , 2016.

International Journal of Scientific Research in Science, Engineering and Technology ( www.ijsrset.com)

128