Volume 8, Issue 6, June 2023 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Deep Learning Implementation for

Poultry Disease Detection and Control
Abuka Emmanuel Ifuchenwuwa1, E.O. Osaghae2, Frederick Duniya Basaky3(PhD)
Professor2, Department of Computer Science,
Federal University, Lokoja, Nigeria

Abstract:-In the subject of Artificial Intelligence (AI) I. INTRODUCTION

known as computer vision, it is possible to teach
computers to comprehend and analyze visual input from The rising population of the world has served to
the outside world, including pictures and videos. highlight the nutritional importance of poultry products,
within the imperative for humanity to rely on animal
It entails the creation of algorithms and systems agriculture. Poultry farming, which comprises a variety of
that can decipher, interpret, and comprehend visual bird species, is the most widely practiced livestock, and its
input in order to carry out operations like image and significance is recognized globally due to the nutrients it
facial identification, object detection and tracking, and provides to the human body (Neethirajan, 2021). Poultry
scene comprehension. Despite the contributions made by products, whether consumed as meat or eggs, provide well-
other academics' work on this subject, there are still balanced, nutrient-dense meals that assist to feed people of
certain difficulties to be resolved. The bottleneck in the all ages in communities all over the world.
image/video processing aspect, among many others, is
the identification, selection, and tracking of bird activity Given the health benefits of chicken products, which
within the poultry farm. This study focuses on the use of have resulted in rising demand on a daily basis, the
cameras in conjunction with specialized AI systems to agriculture industry will consider boosting its efforts to
appropriately analyze flocks for health issues. It also improve production efficiency and quality in order to meet
discusses the integration of AI-assisted technology. market demand. Poultry farmers have benefited from the
Computer vision model was designed using deep rising global demand for poultry products (Caldas et al.,
learning algorithm for monitoring and controlling 2018). Poultry products are a low-cost source of animal
diseases in birds for the purpose of improving the safety protein, which is one of the reasons for the increase
and productivity of the poultry system. This was (Mueller et al., 2018). The poultry industry is divided into
achieved through their captured images from their two categories: egg production and meat production.
droppings (feces) to determine their health state. Data
was collected from two groups of birds; the healthy and The increase in consumption of poultry products is
the unhealthy infected with coccidiosis. The model due to the improvement in human diets that chicken
accuracy was 91% in classifying the healthy and the products have offered (Ren et al., 2020). On an annual basis,
unhealthy feces. By providing an early warning and the manual practice of poultry husbandry often results in
disease incidence prediction, the designed model, when many premature deaths and possible rejection of birds with
put into use, can function as an automatic monitoring one or more defects before they are prepared for
system. This model is to contribute to the advanced consumption. Individual or group assessments of animals on
development of computer vision in recent times and to a constant basis in small-scale agricultural operations can
inform poultry farmers the new trend in the poultry prevent these losses. On a large-scale farm, however, either
system to maximize profit.The quality of poultry group or individual assessment is less practical, regarded as
products will increase as a result of the correct analysis inaccurate and inefficient. This has reignited interest in
of flocks for health issues using cameras integrated to artificial intelligence (AI) techniques, which need computer
specialized systems and Artificial Intelligence (AI) vision systems (Neethirajan, 2021).
technology, which will increase bird survival rates. Recurrent outbreaks of bird diseases have caused
Keywords—Birds, Computer Vision (CV),Convolutional significant setbacks in the poultry sector, resulting in
Neural Network (CNN), Deep Learning (DL), Poultry, financial losses for farmers as well as concerns to human
Droppings, Coccidiosis. health. Some of these diseases can lead to an increase in
human infection, and, as a result, chicken illness has
become a major concern for poultry farmers. Bird flu, for
example, decimated many poultry farms a few years ago,
resulting in significant financial losses for farmers. Manual
bird observation, particularly in large-scale farming,
necessitates a huge number of humans performing checks
on a regular basis, but this is labor-intensive and time-
consuming. As a result, it is necessary to improve
agricultural capability to detect improper behavior in birds

IJISRT23JUN2383 www.ijisrt.com 2926

Volume 8, Issue 6, June 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
in order to preserve good wellbeing without increasing the birds for the purpose of improving the safety and
demand for physical work. This will necessitate the use of productivity of the poultry farming system.
an automated system.  To eliminate the manual method of farm management
with the introduction of Artificial Intelligence (AI).
When it comes to diagnosing diseases in birds, farmers
have always relied on veterinary professionals. Farmers lose D. Significance of the Study
some of their birds as a result of late diagnosis, either while The study of this subject found out that the traditional
waiting for an expert in the field or when an untrained system of managing poultry farms is inefficient, less-
expert is contacted. With the right combination of artificial accurate, leading to economic losses, poor animal welfare,
intelligence (AI) tools, and application of a deep learning premature deaths and production losses.
approach using computer vision for image analysis, we can
easily identify the most common diseases affecting poultry Thus, this study will be highly beneficial to the
birds from acquired images of both droppings (feces) and individual and corporate farmers, researchers, regulatory
captured movements. Research has shown that when the agencies, donor agencies, production lines, and consultants.
correct algorithm is used, automated vision systems are
capable of producing accurate results. As a result, there is a E. Scope of the Study
greater need to focus on using artificial intelligence to It is critical to focus more on using artificial intelligence
reduce disease- related fatalities in chicken production. because automated vision systems have been tested and
shown to be capable of producing results accurately. This
A. Motivation for the study will help decrease the growth in disease-related deaths in
They are frequent outbreak of diseases which affects the poultry farming. Additionally, it emphasizes how computer
poultry industry, causing enormous economic loss to vision can be used in both commercial and research areas to
farmers. Technology developments that are improving in detect diseases. Hence, this study will be based on imagery
accuracy while becoming more affordable hold the key to and advanced deep learning architecture for effectively
solving this problem in the modern era. monitoring and detecting diseases in poultry farms. The
health status of the birds will be determined through their
With the introduction of computer vision, poultry active state and their droppings. These increases the quality
business could benefit from implementing artificial of product, health and survival rates of birds.
intelligence in several elements of farming and
management. F. Limitation of the study
The scope of computer vision covers image and video
B. Statement of the Problem processing for object detection but this work is limited to
In the future, it's anticipated that the consumption of image processing while proposing the video processing for
poultry meat and other poultry-related products would future work. There are two farming system practiced in
continue its previous rapid growth. As a result, chicken poultry, the Battery Cage System and Deep Litter System.
output has considerably expanded worldwide. We will limit our study also to deep litter system. This work
will also be limited to detecting one common poultry
Frequent outbreaks of poultry illnesses are the biggest disease (coccidiosis) were they exist many others.
issue influencing chicken production. (Huang et al., 2019);
(Zhuang et al., 2018); Zhuang& Zhang, 2019), which has II. INTRODUCTION TO COMPUTER VISION IN
often led to significant losses of birds (Rushton et al., 2005). POULTRY FARMING
In order to maintain good wellbeing without increasing the
need for human work, to spot irregularities in chicken There is need to deploy the use of Artificial
behavior and their general health, agricultural capabilities Intelligence (AI) Technology to poultry farming, especially
must be enhanced. Automated systems are required for this. in the African region where the practice is usually 99%
It is necessary to pay greater attention to applying artificial manual. As a result of this practice, farmers often lose birds
intelligence to stop the rise in deaths in poultry farming in large capacity. These birds most times develop diseases
caused by diseases because it has been determined that within the night watch where there is little or no
automated vision systems are capable of generating accurate supervision. Okinda et al., (2020), suggest that “Regular
findings. monitoring to minimize human labor and maintain
affordability is critical to the profitability and productivity
C. Aim and Objectives of commercial chicken production”.
The aim of this research is to monitor and control poultry
diseases in birds using computer vision. In livestock monitoring, human surveillance is no
longer a viable option. As a result, the Precision Livestock
The objectives are; Farming System (PLF) has proven to be a useful alternative
 To create a deep learning-based auto-classification to the problems mentioned by Okinda. These difficulties can
algorithm that could distinguish between healthy and be overcome by implementing efficient automated systems
unhealthy coccidiosis-infected birds. while ensuring acceptable animal welfare. “The Precision
 To propose a computer vision model using CNN Livestock Farming System (PLF) is a stockmen’s support
architecture for monitoring and controlling diseases in system for monitoring various bio-processes and bio-
responses linked to animal welfare, health, and

IJISRT23JUN2383 www.ijisrt.com 2927

Volume 8, Issue 6, June 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
productivity” (Banhazi et al., 2012; Berckmans, 2017; Therefore, the employment of mathematical
Wathes et al., 2008). principles, computer science, and systems integration to
enable image-based automated process control is known as
Because of the advance of technology and the computer vision. Machine learning and deep learning-based
numerous benefits that computer vision systems have systems are the two types of computer vision-based
brought to animal monitoring, PLF with a vision-based systems.
system has become a very important study area. A computer
vision system has the potential of providing non-intrusive Hardware and software make up the two primary
and non-invasive, constant (without distortion), effective, components of computer tracking and monitoring
and objective supervision. There is provision for data technologies for different poultry processes (Natasha et al.,
recording for the purpose of using the data in the future and 2020).
data analysis as well. In addition, computer vision helps
human to reduce tedious and strenuous work which are Hardware components are the camera sensor which
labor-intensive. It has a powerful sensing system that may converts photons into electrical impulses and image
be used to monitor many areas of the farm. “Automated processing board.
tracking platforms can accurately manage the detection and
prediction of abnormal behaviour and poultry diseases” Thermal and infrared (IR) depth-based sensors, as well
(Fang et al., 2020). as visual light-based charge-coupled devices (CCD) and
complementary metal-oxide-semiconductor (CMOS)
The use of these systems in the poultry farming system sensors, are used to capture images of birds in their
allows for data collection and analysis. surroundings in poultry monitoring systems. The digitizer,
also known as the image processing board, turns the visual
A. Historical Perspective to Computer Vision System image into numerical form known as pixel. Other hardware
Computer vision is a field that has developed rapidly in components include wiring, connecting cables and other
the last few decades, although its origins can be found in the tangible components that connect to the computer. Vision
early years of artificial intelligence and computer science. In technology in poultry farming has primarily developed due
the 1950s and 1960s, researchers were already exploring to hardware developments.
how machines could be used to analyze and understand Computer vision typically consist of three essential
images. This was motivated by the desire to create hardware components.
intelligent systems that could process visual information in  Cameras with different lenses appropriate for the setting
the same way that humans do. and the work at hand.
 Lighting units
One of the earliest computer vision systems was the  Mounts that permit a complete view of a farm without
"Perceptron," developed by Frank Rosenblatt in the late interfering with a chicken's regular activities.
1950s. The Perceptron was a simple neural network that
could recognize patterns in images, such as handwritten Software component contains the programs and other
digits. operational data required for the hardware to carry out its
intended purpose.
In the 1970s and 1980s, researchers began to explore
more sophisticated computer vision algorithms, including Software is especially made for data collection and
edge detection, object recognition, and motion analysis. analysis in the agricultural industry, where it must be
These algorithms were used to develop systems that could customized to the specific topic of study. The data
track moving objects, recognize faces, and perform other acquisition software system plays a part in the storage and
visual tasks. selection of high-quality photos (or videos) that are
generated by the cameras. Algorithms appropriate for the
In the 1990s, the development of powerful graphics data and research goals are used by data analytical platforms
processing units (GPUs) and digital cameras helped to to process images (Natasha et al., 2020). The underlying
accelerate the development of computer vision technology. image analysis code in the software manipulates images to
Researchers began to explore new techniques, such as deep produce the desired results.
learning, which uses neural networks to analyze images and
extract features automatically. To obtain images and carry out the desired tasks, many
processing tools have been created and applied. These tools
Today, computer vision system is put to use in include Matlab, ImageJ, and OpenCV, among others.
different applications, including autonomous vehicles, Despite the many benefits of computer vision systems,
surveillance systems, medical imaging, and augmented occlusion issues, background and foreground color contrast,
reality. As technology continues to advance, we can expect and variations in ambient light conditions on farms all have
to see more impressive computer vision applications in the a significant impact on how well any vision system
years to come. performs when monitoring animals.

Notwithstanding, a number of research have been

conducted to address these issues.

IJISRT23JUN2383 www.ijisrt.com 2928

Volume 8, Issue 6, June 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
B. Effects of Diseases on Poultry Farms of computer vision can identify diseases as quickly as
Diseases can have significant effects on poultry farms, possible.
both in terms of animal welfare and economic impact. Some
of the common effects of diseases on poultry farms include: C. Cossidiosis and its Effect on Poultry Farms
 Mortality: Diseases can cause a significant number of Coccidiosis is a common disease that affects poultry,
poultry to die, which can result in financial losses for the particularly chickens and turkeys. It is caused by a group of
farm. High mortality rates can also reduce the overall parasites called coccidia, which belong to the genus
productivity of the farm and decrease the supply of Eimeria. These parasites infect the intestinal lining of
poultry products. poultry and can cause a range of symptoms, including
 Reduced growth rate: Some diseases can affect the diarrhea, weight loss, and anemia. In severe cases,
growth rate of poultry, which can result in reduced weight coccidiosis can lead to death. The disease affects both
gain and slower development. This can increase the time commercial broiler chickens and laying hens, and can cause
and resources needed to raise the poultry to market significant economic losses for poultry farmers.
weight, which can also result in economic losses.
The coccidia parasites are transmitted through the
 Decreased egg production: Diseases that affect the feces of infected birds, and poultry can become infected by
reproductive system can reduce the number of eggs ingesting contaminated feed or water. The parasites multiply
produced by poultry. This can decrease the supply of eggs inside the intestinal lining and cause damage to the tissue,
for sale, resulting in a reduction in revenue for the farm. leading to inflammation and diarrhea. In severe cases, the
 Spread of disease: Diseases can spread quickly among parasites can cause the intestinal lining to become so
poultry, particularly in crowded conditions. If not damaged that it becomes prone to bacterial infections.
addressed promptly, disease outbreaks can lead to the
entire flock becoming infected and significant economic Poultry that are infected with coccidia may also
losses. experience weight loss and anemia due to the loss of
 Cost of treatment: Treating diseases in poultry can be nutrients and blood resulting from the diarrhea. In addition,
expensive, particularly if large numbers of birds are the immune system of infected birds may be compromised,
affected. The cost of veterinary services, medication, and making them more susceptible to other diseases.
other treatment options can add up quickly and cut into
the farm's profits. To prevent coccidiosis, it is important to maintain
 Public health concerns: Some poultry diseases can pose good hygiene practices in poultry farms, including regularly
a risk to human health. For example, avian influenza can cleaning and disinfecting feeders and watering systems, and
be transmitted to humans, leading to serious illness or disposing of feces properly. In addition, some farmers may
evendeath. This can have significant public health use medications or vaccines to help prevent or treat
implications and result in the loss of consumer confidence coccidiosis. It is important for poultry farmers to be aware
in poultry products. of the signs and symptoms of coccidiosis, and to take
appropriate preventative measures to reduce the risk of
To mitigate the effects of diseases on poultry farms, it outbreaks. This can include regular monitoring of birds for
is important for farmers to pay attention to the use of signs of illness, and implementing a biosecurity plan to
artificial intelligence tools to control outbreak of diseases minimize the risk of disease transmission. Since human
thus maintaining high standard of biosecurity were the use monitoring may not be effective, hence the need for a vision
based system that can guarantee 24hrs monitoring.

Fig. 1(a): Sample pictures of Coccidiossis

Fig. 1(b): Sample pictures of Coccidiossis

IJISRT23JUN2383 www.ijisrt.com 2929

 Volume 8, Issue 6, June 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
III. METHODOLOGY The birds were expected to be tagged using Radio-
Frequency Identification (RFID) but due to the nature of this
When coccidiosis outbreaks occur in poultry farms, work, using tracking system will not be required because in
both the birds and the farmers suffer severe economic the practice of poultry, entire birds are expected to be
losses. In spite of the slow recuperation and reestablishment treated alike after a symptom of diseases is discovered at
period following treatment, there is a sudden and significant early stage.
decline in production figures. Some flocks never reach their
maximum level of productivity or recovery. The fact that A. Tools and Method
treatment cannot stop the financial losses is thus widely The model was generated through Google colab
acknowledged. The sole option, then, is disease prevention, environment using python v3, with embedded Tensor flow
which is widely accepted throughout the poultry industry. packages, (using keras library function) and running the
training set in a 16 GB RAM computer, running Microsoft
As a result, there is a need for technologically based windows 10, of 250g SSD Hard disk.
solutions that are efficient and reasonably priced. Yet,
developing an effective and sustainable prevention and Dropping images were collected from Nathliz Organic
control program against the disease is difficult. Poultry Farm, Lokoja using mobile phone camera of
1600X1200 (4:3) resolutions in a Joint Photographic Group
This work adopts computer vision approach for (JPG) format to collect the data needed. Images for their
tracking and monitoring the wellbeing of birds at droppings were collected and distributed into 2 class labels,
NathlizOraganic Poultry Farm, Lokoja. A total of 500 birds namely; Healthy with 2056 records and Unhealthy
(broilers) kept from day old till 8 weeks old were observed. (Coccidiosis) with 2133 records.
Coccidiosis is a common disease within birds and always
surface between 3-6 weeks. The active states of the birds This work will also adopt some steps in CRISP DM
were determined from captured images. But we focus on (Cross Industry Standard Process) with the following
their droppings (feces) as earlier mentioned. The target of process;
this model is to detect the presence of coccidiosis through  Data interpretation
the dropping of the birds. The practice of Poultry, when  Data preparation
droppings (feces) from birds are detected to be changed  Modeling
from healthy to unhealthy state, it shows through their feces.  Evaluation/classification
Brownish feces are indication of the presence of coccidiosis.
B. Dataset
Camera phone was used as image capturing The picture dataset were gotten from Nathliz Organic
component to acquire images (droppings) from the poultry Poultry Farm, Lokoja, Nigeria and distributed in two classes
farm. The image processing board converts the visual image namely, healthy and unhealthy (coccidiosis).
into numerical form (pixels); the output component
produced the underlying image analysis code for image
manipulation.

Fig 2.Sample of healthy and droppingUnhealthy Dropping (Coccidiosis)

IV. SYSTEM DESIGN AND IMPLEMENTATION artificial neural networks are capable of processing picture
data, CNNs have showed impressive performance and
In this study, we develop a solution which is higher accuracy.
Convolutional Neural Network (CNN) based. The CNN
architecture was designed from scratch then trained the CNN Architecture: The proposed CNN architecture
model to detect the presence of a disease (Cossidiosis) and involves stacking of multi convolution layers as sampled in
return a healthy or unhealthy output. Fig 3.

Most activities requiring picture data, or datasets, are A CNN ignores batch size and accepts tensors of
performed using CNNs. Each image has pixel information shape (image height, image width, and color channels) as
that can be expressed numerically. A CNN receives this input. In the first layer, a specified image size and height of
numerical representation as input. While conventional 25X25, RGB was fed to the stack of convolution layer as

IJISRT23JUN2383 www.ijisrt.com 2930

Volume 8, Issue 6, June 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
input. The convolutional layer uses first 32 and then 64 During the training of the model, normal stochastic
filters with a 3×3 kernel as a filter followed by the max- gradient descent was used to minimize the error, and in
pooling layer which performs over a 2x2 pixel window. evaluation, we leverage log loss and accuracy as the metrics.

These layers combine to form a single block, which Because the problems proposed to solve are discreet in
we continually apply by raising the depth of filter from nature, we used binary crossentropy loss function which
16,16, 32, 32, 128 for the full convolution block. Before we seems to fit the problem with the log loss as an evaluation
built a fully meshed network, we replace the 3×3 matrix per metric.
image, with a vector of 512 elements (4*4*32), without
losing any information before passing through two dense A. Objectives of the Design
layers. Using our created collection of fecal images from
Nathliz Organic Poultry Farm, Lokoja, we seek to establish
The same padding was used in each block to keep the an automated computer vision model for early identification
height and breadth shapes of the output feature maps and classification of diseases in birds.
consistent with the input features. Weight initialization for
all blocks takes shape uniformity into account. ReLU B. Architectural Model Structure of Convolutional Neural
activation was used for all layers before applying the Network
sigmoid function at the last layer. Sigmoid activation The convolutional neural network's task is to retain
function was used since the problem is not a multi-class elements necessary for a good prediction while compressing
classification and to be able to have a good prediction of the images into a more manageable format.
either healthy or unhealthy feces.
This is essential for developing an architecture that can
scale to large datasets and learn features.

Fig. 3: Architectural structure of CNN

C. Result and Analysis each epoch having two steps, given a total of 20X2 (40
The results after assessing the model's performance are iterations) for a complete training process. From the graph
shown in this section. Brief descriptions of the dataset and in Fig. 4(a) and Fig. 4(b), it can be seen that the accuracy of
the partitioning of the data into the training set, validation validation keeps increasing and converges at the 20th epoch.
set, and test set are provided in Table 1. Default learning rate of 0.001 was used while the maximum
learning rate was attained at epoch 14.
We started by showing how the computer vision
system components can be designed or constructed before Total raw dataset collected was 4189 images for both
developing the CNN model for our diseases classification. healthy and unhealthy droppings, 2056 for the healthy and
To finish the training process, we performed 20 epochs with 2133 unhealthy (Coccidiosis) as shown in Table 2.

Table 1: Training, Validation and Testing sets (Dataset Splitting)

S/N Class No. of Dataset Training Validation Test
1. Healthy 2056 1561 495 495
2. Unhealthy (Coccidiosis) 2133 1712 421 421
Total 4189 3273 916 916

As a result of data cleaning performed on the dataset, photos using a variety of techniques to increase the size of
the number of data collected from the poultry farm was the dataset. Our model uses the augmentation techniques of
reduced, necessitating the need for data augmentation. image flipping, image cropping, and padding. Image
During the training, To minimize over-fitting and to enable resolution was also lowered to cut down on training time.
the model to learn from unseen datasets, we enhanced the

IJISRT23JUN2383 www.ijisrt.com 2931

Volume 8, Issue 6, June 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Since CNN needs a constant input dimensionality of model's predictions. Losses will be at their lowest when
the data to train the optimal model, resizing the images has model predictions are closest to the actual values and at
the advantage of achieving a suitable resolution. The model their worst when forecasts are too distant from the starting
was not hyper trained hence the model used the default values. The evaluation used for the model is the input
learning rate. The loss function indicates the accuracy of evaluation function.evaluate with accuracy of 91%.

Table 2: Performance Result For Cnn Model

Log Loss 0.1521
Validation Accuracy 0.913
Validation Loss 0.199

Fig 4(a): Model Accuracy Graph

Fig 4(b): Model Loss Graph

IJISRT23JUN2383 www.ijisrt.com 2932

Volume 8, Issue 6, June 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165

 Application Algorithm for Proposed Model

Start

Upload
Droppings

Process Acquired
Droppings

Is YES
Coccidiossis Unhealthy
detected?

NO

Isolate
Healthy

End

Fig. 5: Flowchart of proposed model

IJISRT23JUN2383 www.ijisrt.com 2933

Volume 8, Issue 6, June 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
D. Application Designed Interface
An application was designed from the designed model showing user interface were pictures of chicken droppings can be
uploaded to further classify the image either healthy or unhealthy. A pictorial view is shown below;

 User interface to upload faecal images

Fig. 6(a): User interface for uploading fecal picture

Fig 6 (b)User interface showing unhealthy faeces with the presence of coccidiosis

Fig. 6(c): User interface showing healthy faeces without the presence of coccidiosis

IJISRT23JUN2383 www.ijisrt.com 2934

Volume 8, Issue 6, June 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
V. SUMMARY REFERENCES

Computers technologies (such as computer vision) can [1.] Athiraja, A., &Vijayakumar, P. (2021). Banana
be deployed in monitoring the growth processes of birds in disease diagnosis using computer vision and machine
real time in an automated fashion which is less time learning methods. Journal of Ambient Intelligence
consuming and effective. Identifying the length and and Humanized Computing, 12(6), 6537–6556. DOI:
frequency of postural shifts in birds over time can aid in the https://doi.org/10.1007/s12652-020-02273-8.
prevention of sickness outbreaks that could put their health [2.] Aydin, A., Cangar, O., ErenOzcan, S., Bahr, C.
at risk. However, the designed model using Convolutional &Berckmans, D. (2010). Application of a fully
Neural Network was built from scratch with 91% accuracy automatic analysis tool to assess the activity of
indicating greater level of performance. broiler chickens with different gait scores, Computers
and Electronic in Agriculture, 73(2), 194-199, DOI:
A. Review of Contributions and Achievements 10.1016/j.compag.2010.05.004.
It is important to detect diseases in birds timely and [3.] Ayim-Akonor, M., Krumkamp, R., May, J.,
allow veterinarians take necessary actions as expert in the &Mertens, E. (2020). Understanding attitude,
field thereby preventing loses and as well maintain good practices and knowledge of zoonotic infectious
animal welfare status. Thus, the model can be applied as an disease risks among poultry farmers in Ghana.
automated vision monitoring system for the whole life span Veterinary Medicine and Science, 6(3), 631–638.
of birds reducing the need for frequent visits from DOI: https://doi.org/10.1002/vms3.257.
veterinarians. The application of related works has helped in [4.] Bi, M., Zhang, T., Zhuang, X., & Jiao, P. R. (2018).
the loss of birds as early warning signs are observed through Recognition method of sick yellow feather chicken
the use of Artificial Intelligence. based on head features. Transactions of the Chinese
Society for Agricultural Machinery, 49(1), 51-57.
B. Recommendations DOI: https://doi.org/10.6041/j.issn.1000-
Monitoring birds’ behavior allows farmers to identify 1298.2018.01.006.
pathological changes early and discover factors threatening [5.] Caldas, J.V., Hilton, K., Boonsinchai, N., England,
the health of birds in advance. Poultry farming are J.A., Mauromoustakos, A., & Coon, C.N. (2018).
widespread venture both on small and large scales. Research Dynamics of nutrient utilization, heat production, and
has proven that manual observation of these birds is time body composition in broiler breeder hens during egg
consuming and labor intensive. production. Poul. Sci., 97(8), 2845–2853.
Therefore, the proposed model will be recommended https://doi.org/10.3382/ps/pey133.
for implementation to determine the effectiveness of the [6.] Chmiel, M. Słowiński, M., &Dasiewicz, K. (2011).
model in real-world scenario for future works which will Application of computer vision systems for
greatly be of immense help to poultry farmers in managing estimation of fat content in poultry meat,” Food
their farms effectively. Control, 22(8), 1424–1427, DOI:
10.1016/j.foodcont.2011.03.002.
C. Areas for Future Research [7.] Fang, C., Huang, J., Cuan, K., Zhuang, X. and Zhang,
Several diseases exist that have impact on poultry T. (2020). “Comparative study on poultry target
farming; hence, for further research, we anticipate training tracking algorithms based on a deep regression
various illnesses applying computer vision systems in real- network,” Biosyst. Eng., vol. 190, February 176–183.
time practice, with the inclusion of video processing DOI: 10.1016/j.biosystemseng.2019.12.002.
method, to ascertain the health and welfare of the birds. [8.] Ferentinos, K.P. (2018) “Deep learning models for
plant disease detection and diagnosis,” Computers
VI. CONCLUSION and Electronics in Agriculture, 145, February, 311–
318. DOI:
For several decades, computer vision research has https://doi.org/10.1016/j.compag.2018.01.009.
been attempting to close the gap by developing automated [9.] Guanjun, B., Mimi, J., Yi, X., Shibo, C., &Qinghua,
systems that can analyze images for decision-making using Y. (2019). Cracked egg recognition based on
computers, but this work has not yet produced any tangible machine vision,” Comput. Electron. Agric., 158
benefits. (March), 159–166.
https://doi.org/10.1016/j.compag.2019.01.005.
CNN model was designed to learn the hidden pattern [10.] Huang, J., Wafng, W., & Zhang, T. (2019). Method
from the several fecal images acquired through the dataset. for detecting avian influenza disease of chickens
The deep learning algorithm designed was able to based on sound analysis. Biosystems Engineering,
predict the state of the birds either healthy or unhealthy 180, 16-24. DOI:
(coccidissis) were the results obtained show the proposed https://doi.org/10.1016/j.biosystemseng.2019.01.015.
model achieved up to 91% accuracy. [11.] Kashiha, M.A., Bahr, C., Vranken, E., Hong, S.
&Berckmans, D. (2014). Monitoring system to detect
problems in broiler houses based on image
processing,” in Proc. International Conference of

IJISRT23JUN2383 www.ijisrt.com 2935

Volume 8, Issue 6, June 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Agricultural Engineering, Zurich, Switzerland, (pp. [22.] Sadeghi, M., Banakar, A., Khazaee, M. &Soleimani,
1-7). DOI: 10.1016/j.compag.2016.03.011. M. R. (2015). Intelligent Procedure for the Detection
[12.] Legnardi, M., Tucciarone, C. M., &Franzo, G. and Classification of Chickens Infected by
(2020). Infectious Bronchitis Virus Evolution, Clostridium perfringensBased on Their Vocalization.
Diagnosis and Control. Veterinary Sciences, 7(2), 79. Braz. J. Poult. Sci., 17, 537-544. DOI:
DOI: https://doi.org/10.3390/vetsci7020079. https://doi.org/10.1590/1516-635X1704537-544
[13.] Mortensen, A.K., Lisouski, P. &Ahrendt, P. (2016). [23.] Shao, H., Pu, J., & Mu, J. (2021). Pig-Posture
Weight prediction of broiler chickens using 3D Recognition Based on Computer Vision : Dataset and
computer vision, Computers and Electronics in Exploration. Animals, 11(5), 1295. DOI:
Agriculture, 123, 319-326. DOI: https://doi.org/10.3390/ani11051295.
10.1016/j.compag.2016.03.011 [24.] Singh, M., Kumar, R., Tandon, D., Sood, P., &
[14.] Mueller, S., Kreuzer, M., Siegrist, M., Mannale, K., Sharma, M. (2020). Artificial Intelligence and IoT
Messikommer, R. E., &Gangnat, I. D. (2018). based Monitoring of Poultry Health: A Review. 2020
Carcass and meat quality of dual-purpose chickens IEEE International Conference on Communication,
(Lohmann Dual, Belgian Malines, Schweizerhuhn) in Networks and Satellite, Proceedings of IEEE
comparison to broiler and layer chicken types. Poul. International Conference on Communication,
Sci. 97(9), 3325–3336. DOI: Networks and Satellite (Comnetsat) 2020, pp 50–54.
https://doi.org/10.3382/ps/pey172. DOI:
[15.] Nasiri A, Omid M., & Taheri-Garavand A . (2020). https://doi.org/10.1109/Comnetsat50391.2020.93289
An automatic sorting system for unwashed eggs 30.
using deep learning. Journal of Food Engineering, [25.] Wathes, C.M. Kristensen, H.H., Aerts J-M., &
283, 1-9. Berckmans, D. (2008). Is precision livestock farming
https://doi.org/10.1016/j.jfoodeng.2020.110036 an engineer's daydream or nightmare, an animal's
[16.] Neethirajan, S. (2021). Automated tracking systems friend or foe, and a farmer's panacea or pitfall?
for the assessment of farmed poultry. Preprint, (pp1– Computers and Electronics in Agriculture, 64(1), 2-
13). DOI: 10. DOI:10.1016/j.compag.2008.05.005.
https://doi.org/10.20944/preprints202105.0364 [26.] Ye, Chang-wen, Yu, Zhen-wei, Kang, R., Yousaf, K.,
[17.] Noldus, L.P.J.J., Spink, A.J., &Tegelenbosch, R.A.J. Qi, C. Chen, Kun-jie, & Huang, Yu-ping. (2020). An
(2002). Computerized video tracking, movement experimental study of stunned state detection for
analysis and behavior recognition in insects. broiler chickens using an improved convolution
Computers and Electronics inAgriculture, 35(2-3), neural network algorithm. Computers and Electronics
201-227. DOI: in Agriculture, 170, 105284. DOI:
doi.org/10.1016/S0168%2D1699(02)00019%2D4 https://doi.org/10.1016/j.compag.2020.105284.
[18.] NurSyazarin Natasha Abd Aziz, Salwani Mohd Daud [27.] Zhuang, X., Bi, M., Guo, J., Wu, S. & Zhang, T.
Rudzidatul Akmam Dziyauddin, MohamadZulkefli (2018). Development of an early warning algorithm
Adam &AzizulAzizan (2020): A Review on to detect sick broilers. Computers and Electronics in
Computer Vision Technology for Monitoring Poultry Agriculture, 144(1), 102–113.
Farm – Application, Hardware, and Software. IEEE https://doi.org/10.1016/j.compag.2017.11.032.
Access, 9, 12431-12445. DOI: [28.] Zhuang, X., & Zhang, T. (2019). Detection of sick
http://dx.doi.org/10.1109/ACCESS.2020.3047818 broilers by digital image processing and deep
[19.] Ojukwu, C.C., Feng, Y., Jia, G., Zhao, H., & Tan, H. learning. Biosystems Engineering. 179, 106-116.
(2020). Development of a computer vision system to DOI:
detect inactivity in group-housed pigs. International https://doi.org/10.1016/j.biosystemseng.2019.01.003.
Journal of Agricultural and Biological Engineering,
13(1), 42–46. DOI:
https://doi.org/10.25165/j.ijabe.20201301.5030
[20.] Okinda, C., Nyalala, I., &Korohou, T. (2020). A
review on computer vision systems in monitoring of
poultry: A welfare perspective. Artificial Intelligence
in Agriculture, 4, 2020, 184-208. DOI:
https://www.sciencedirect.com/science/article/pii/S25
89721720300258
[21.] Ren, G., Lin, T., Ying, Y., Chowdhary, G., & Ting,
K. C. (2020). Agricultural robotics research
applicable to poultry production : A review.
Computers and Electronics in Agriculture, 169
(December 2019), 105216. DOI:
https://doi.org/10.1016/j.compag.2020.105216

IJISRT23JUN2383 www.ijisrt.com 2936