2024 Second International Conference on Advances in Information Technology (ICAIT-2024)

Real-time Animal Identification and Alert

System using IoT and Deep Learning

Manujakshi B C Shashidhar T M Ravikiran H N

Department of Computer Department of Computer Department of Electronics
Science and Engineering, Science and Engineering, and Communication
Faculty of Engineering and Acharya Institute of Engineering,
Technology, JAIN (Deemed- Technology, Bengaluru, Adichunchanagiri University,
2024 Second International Conference on Advances in Information Technology (ICAIT) | 979-8-3503-8386-7/24/$31.00 ©2024 IEEE | DOI: 10.1109/ICAIT61638.2024.10690384

to-be-University), Bengaluru, Karnataka, India Nagamangala, Mandya,

Karnataka, India Karnataka, India

Asha R Rajeev Bilagi

Department of Electronics Department of Computer
and Communication Science and Engineering,
Engineering, Vidya Vikas Acharya Institute of
Institute of Engineering and Technology, Bengaluru,
Technology, Mysore, Karnataka, India
Karnataka, India

Abstract The amount of traffic accidents that result in fatalities

warning signs have been put up to lessen the incidence of animal-
and injuries is one of the most urgent problems facing the world vehicle crashes. These collisions are caused by things like
today. Many tragic incidents are caused by collisions between inattentive driving, a lack of awareness, and a failure to notice
humans and animals. Worldwide, there are startlingly many signs. The development of animal detection applications is
automobile accidents—one every few seconds. Our primary essential to the resolution of practical issues. The majority of
objective is to decrease the probability of traffic accidents caused these apps rely on the ability to recognize animals in pictures or
by animals by identifying them and informing cars of their videos. Thus, our goals are to save animals, lower the global
presence. We suggest using IoT technology to create a detection death toll, and lessen the number of accidents-related wild animal
and warning system that can significantly lower animal-related deaths. Reducing animal-vehicle collisions is the primary goal of
incidents. The purpose of this warning system is to alert cars on
the roadway as soon as a detection event takes place by means of
animal detection. Our goal is to create an accurate and efficient
a signal. Our solution is based on Internet of Things technology, system that may be deployed inside cars or at regular intervals
with the Raspberry Pi3 Model B acting as the main component along highways to alert drivers to the presence of animals and
for car warning and animal identification. Real-time request that they slow down
photographs and videos of animal movements are captured by
the Pi camera, which is equipped with a Raspberry Pi, through A. Internet of Things
the use of sensors to identify impediments. After that, it alerts any A network of interconnected computing devices, digital and
vehicles and pedestrians using the woodland road. Additionally, mechanical machines, objects, animals, or people is referred to
it marks particular distances at night with lights. In the end, the
as the Internet of Things (IoT). Each of these devices is given a
document's primary goal is to safeguard numerous people from
possible harm and to preserve animals from mishaps. In order to unique identity and is capable of transmitting data over the
do this, the YOLO (You Only Look Once) algorithm was created. network without the need for human intervention, such as in
It processes approximately 45 frames per second and examines human-computer interaction. Internet of Things-based
the complete picture in order to determine if objects are present technologies have the potential to revolutionize daily life and
or not. offer enhanced services in the near future.
Keywords: YOLO model, IoT, Deep Learning, Raspberry Pi Technological advancements in medical, energy, agriculture,
gene therapy, smart cities, and smart homes are a few areas
I. INTRODUCTION where they are strongly present. Currently, the network is home
to approximately 9 billion items. This number is predicted to
The amount of road traffic accident-related deaths and increase to as much as 20 billion with the advent of the Internet.
injuries is a pressing global issue that we are currently
facing. One of the primary causes of these regrettable A. IoT relies on four principal components
incidents is collisions between humans and animals. Around i. Low-power embedded systems: When designing electronic
the world, there are road accidents every fifteen minutes. systems, these systems put great performance and low battery
Serious issues arise from human-animal conflicts, which usage first.
frequently cause crop damage, human casualties, and
human-killed animals. Specifically, died while driving
through forests. Accidents still happen even if different ii. Cloud computing: Reliable storage servers are necessary due

979-8-3503-8386-7/24/$31.00 ©2024 IEEE

Authorized licensed use limited to: Yuan Ze University. Downloaded on October 28,2024 at 17:01:29 UTC from IEEE Xplore. Restrictions apply.
 2024 Second International Conference on Advances in Information Technology (ICAIT-2024)
to the high volume of data created by Internet of Things features and results of the DETER-R (Deforestation Protection
devices. Processing and analyzing the data to generate Near Real-Time) system, an automated deforestation detection
insights—such as locating electrical faults or system system centered on the Brazilian Amazon, based on Synthetic
errors—is made possible in part by cloud computing. Aperture Radar (SAR) are presented in this article. [2]
Deforestation and natural disasters (e.g. forest fires, increased
gas emissions) pose several serious challenges to forest
iii. Large data availability: Since real-time sensors are
ecosystems. This study proposes the use of analog and digital
essential to the Internet of Things, their widespread use in a
sensors, signal analysis techniques, and Raspberry Pi Model 3 as
wide range of applications will inevitably result in a rise in
the basis for an intelligent forest environmental monitoring
large data.
system. In addition to their great potential to slow down human-
induced climate change, forests [3] provide many other benefits
iv. Network connectivity: Every physical thing requires an
to society. However, in the 21st century, 98 forest carbon sinks
IP address-based Internet connection in order to
could be seriously threatened by climate change-related
communicate effectively. To represent real objects,
challenges. Here we summarize current knowledge on climate-
researchers must investigate alternate naming systems
driven threats to forest stability from fire, drought, biological
because the existing system runs out of IP addresses as the
agents, and other disturbances99. Recent developments,
number of devices rises.
including in [4], are near-real-time (NRT) forest monitoring and
B. Deep Learning warning systems that use remote sensing data to support the
management and protection of forest resources. Local response
teams tasked with assessing and mitigating threats to tropical
Within the study of artificial intelligence and machine
forest ecosystems can benefit greatly from rapid and accurate
learning, deep learning is a subset of machine learning. It
detection of deforestation-related activities using satellite and
deals with networks that are capable of autonomous learning
airborne remote sensing. This study [5] proposes a versatile,
from unlabeled or unstructured data. The functioning of
easily deployable, and scalable wildfire detection and
deep neural networks, deep learning, and deep neural
management system based on sensor networks (SNs) that can be
learning is strikingly similar to that of the human brain. Deep
used in a variety of environments and regions. SNs are deployed
learning algorithms are characterized by their capacity to
in hazardous and sensitive locations to monitor the environment
continually examine structured data and deduce human
and provide relevant data for analysis. Actively detecting and
thinking by using a logical framework. Deep learning makes
managing wildfires can be challenging in certain situations,
use of a multilayer algorithm known as a neural network to
especially in isolated and difficult-to-access areas or within fire
carry out this purpose.
hotspots.

III. METHODOLOGY

The suggested system architecture design is displayed in Figure

2. Our system's accuracy is mostly dependent on the ultrasonic
sensor's capacity to recognize objects in its immediate
surroundings, including people and animals. Upon identifying an
animal, the sensor triggers the camera to record video. After
being encoded into frames, the captured video is supplied to the
model. YOLO is an efficient and accurate image detection
algorithm that is used to train the model on the animal dataset.
Fig. 1: Neural Network After identifying an animal in the picture, the model classifies
the animal and outputs its position, which is subsequently entered
A neural network's fundamental architecture, comprising into the database using the Flask backend.
the input and output layers, is depicted in Fig. 1. A neural
network contains one input and one corresponding output In addition, a signal is transmitted to turn on the buzzer and LED
for each input layer.
lights as a warning system when an object is detected. The driver
warning application's backend continuously searches the
The fundamental building blocks of a neural network are
the neurons within it. These networks are made up of database for further new asset items before sending the
hidden neurons in the network's hidden layer, input information to the system's frontend, which plots it on a map. In
neurons that correspond to the input layer, and output order to prevent a potential accident with the wild animal, the
neurons that correspond to the output layer. To generate driver is alerted by an auditory warning and receives a map
the appropriate output, the hidden neurons apply weights showing the animal's location. This specific technique has the
to the input and then send it via an activation function. ability to greatly lower the frequency of collisions between
They route the data after applying a nonlinear change to it. humans and animals, hence reducing the number of traffic
accidents brought on by wild animal interactions.
II. RELATED WORK
The research work mentioned in [1] includes the main
979-8-3503-8386-7/24/$31.00 ©2024 IEEE

Authorized licensed use limited to: Yuan Ze University. Downloaded on October 28,2024 at 17:01:29 UTC from IEEE Xplore. Restrictions apply.
 2024 Second International Conference on Advances in Information Technology (ICAIT-2024)

Fig. 2: Proposed System Design

IV. RESULTS

In this study, we employ an ultrasonic sensor and a

Raspberry Pi device with a Pi camera to detect animals, as
illustrated in Figures 3, 4, and 5. The major goal is to identify
wildlife and alert approaching motorists to reduce the
number of wildlife-related accidents on forest routes. The
primary warning systems for cars include a driver alert app,
LED lights, and a buzzer. DC motors are also used by it for
particular rotational tasks. Every component of the system is Fig. 3: Animal class Elephant Object Detection
thoughtfully positioned at key points along the woodland
trail. An ultrasonic sensor recognizes the presence of an
animal as it gets closer to the road and triggers the Pi camera.
In order to classify the animal, the camera then captures a
live image of it and feeds it to a deep learning model. Both
an LED light and an audio buzzer sound when the sensor
triggers an alarm, warning other cars in the area. Data about
the locations and classification of animals is kept in a
database. An alert program keeps a constant eye on this data,
checking every second for new entries. The animal's location
is indicated on the map, and an audible alert is produced by
the map display. In the end, this method lowers the number
of accidents on forest routes and the harm done to both
wildlife and cars by encouraging cars to slow down and give
animals enough space to cross or move across the road.

Fig. 4: Animal class Tiger Object Detection

979-8-3503-8386-7/24/$31.00 ©2024 IEEE

Authorized licensed use limited to: Yuan Ze University. Downloaded on October 28,2024 at 17:01:29 UTC from IEEE Xplore. Restrictions apply.
 2024 Second International Conference on Advances in Information Technology (ICAIT-2024)
[4] John Musinsky, Karyn Tabor et.al. Conservation impacts of a near
real-time forest monitoring and alert system for the tropics,2018, Remote
Sensing in Ecology and Conservation.

[5] Sharma, B.S. Sohi, and S. Chandra, SN Based Forest Fire Detection
and Early Warning System, International Journal of Innovative
Technology and Exploring Engineering (IJITEE). ISSN: 2278-3075,
Volume-8 Issue-9, July 2019

Fig. 5: A Driver Alert Application Map

V.CONCLUSION
Machine learning techniques, particularly deep learning,
have become indispensable in many facets of our life in
today's technologically advanced society. Numerous
algorithms and applications employ these techniques. This
work combines deep learning with IoT to propose a novel
method for detecting wildlife. The primary objective is to
recognize wildlife and promptly alert passing cars on forest
roads. It is well known that forest road accidents frequently
result in injury or death to animals and negatively affect
endangered species. The Raspberry Pi3 Model B is used in
our project to identify items, such animals, and alert passing
cars to their presence. We accomplish this by detecting
obstructions using an ultrasonic sensor. The Pi camera
detects obstacles and turns on to capture real-time images of
the animals. Deep learning techniques are then used to
examine these photos; these methods are kept in a database.
The car on the forest road is then alerted to possible danger.
Future developments might see the system integrated into
automobiles, its detecting range expanded, and its
autonomous speed lowered to prevent crashes. Our goal is
for our study to help prevent accidents and conserve animals.
REFERENCES
[1]. Doblas, J.,Reis, M.S., Belluzzo A.P, Quadros,C.B, Moraes,
D.R.V, Almeida, C.A., Maurano, L.E.P, Carvalho, A.F.A.,
Sant’Anna, S.J.S.; Shimabukuro, Y.E. DETER-R: An Operational
Near-Real Time Tropical Forest Disturbance Warning System Based
on Sentinel-1 Time Series Analysis. Remote Sens. 2022, 14, 3658.
https://doi.org/10.3390/rs14153658

[2] Garcia, M., & Kim, S. (2018). IoT-Based Forest Monitoring for
Earthquake Resilience. Forestry Research Journal, 22(3), 245-257.

[3] Johnson, A., Davis, B., &Smith, J. (2019). Climate-Driven

Predictive Analytics for Forest Health. International Journal of
Environmental Management, 36(5), 621-635.

979-8-3503-8386-7/24/$31.00 ©2024 IEEE

Authorized licensed use limited to: Yuan Ze University. Downloaded on October 28,2024 at 17:01:29 UTC from IEEE Xplore. Restrictions apply.