Unlocking Insights: Object Detection in Aquarium

Himanshu S Bardhiya1 Dr. Rajendra Rewatkar2 Mutyala Pavitra3

Artificial Intelligence & Data Science,FEAT Associate Professor Computer Science & Design,FEAT
Datta Meghe Institute of Higher Education Department of Biomedical Datta Meghe Institute of Higher Education
and Research Engineering,FEAT and Research
Sawangi, India Datta Meghe Institute of Higher Education Sawangi, India
[email protected] and Research [email protected]
Sawangi, India
[email protected]

Abstract— Paper presents monitoring and controlling of implement the necessary adjustments to keep it in optimal
temperature, pH levels, illumination, and feeding schedules condition.
using sensors. It will automatically modify settings to maintain
the ideal aquarium atmosphere for the fish and plants by Every operation in the aquarium, such as feeding and
utilizing IoT. This makes aquarium maintenance more watering, replenishment, monitoring illumination,
convenient and environmentally beneficial. The system intends temperature and pH regulation is done
to enhance the management and monitoring of several factors automatically. Because the aquarium management process is
in the aquarium environment, including temperature, pH automated, less manual labor will be needed to maintain
levels, lighting, and feeding schedules, by utilizing Internet of
tanks.
Things technology. Real-time data gathering as well as analysis
enable accurate changes and automatic reactions to sustain The challenge of keeping an aquarium is increased by the
ideal circumstances for aquatic life using sensor networks with requirement that the fish be fed twice a day, a task that
cloud connections. The use of this kind of technology has the necessitates the owner coming up to the fish tank in person.
potential to increase productivity, decrease the need for
manual intervention, and promote environmentally friendly
Sometimes, when on vacation, the owner is unable to
aquarium management methods. Real-time data analysis maintain control over the tank and feed the fish. Compared to
makes use of sensors and cloud connectivity which provide existing market solutions, this model will be more effective
automatic modifications for ideal conditions of aquatic life, and cost less. The target audience for the model would be
hence it offers more efficiency and sustainability in aquarium those who want to maintain a fish aquarium in their houses
care. or places of business. But, they are not able to take care and
need to take help of their neighbours while living away or
Keywords— Aquarium, Design, Internet of things, busy with their work. The aim of this proposed model is to
Automated, Detection, Dataset
create a system for automatically caring for fish using smart
I. INTRODUCTION sensors and IoT technology. [1]

Pet ownership has steadily increased during the past 20 II. LITERATURE REVIEW
years. Nowadays, dogs and cats are the most popular pets, According to J. Eksteen, L. Coetzee, et al [2], the advent
followed by freshwater fish. Keeping a fish tank maintained of the notion of the Internet of Things resulted in a profound
may be an extremely challenging undertaking in and of itself. change. In recent years, the Internet of Objects (IoT) has
There are several things needed to be done when feeding or advanced beyond computers to people to objects, opening up
cleaning the aquarium will be done. The power head and air a vast array of applications and services. Many industries are
pump on the aquarium must be turned off. Operator must embracing IoT concepts for their purposes and providing
hand-feed the tank and restart the air pump later. Under the advice on how to put them into practice. Every technological
current setup, all equipment as well as electrical switches industry will see globalization as IoT continues to progress.
must be used to manually turn on and off appliances like IoT has a bright future ahead of it as it continues to flourish
lights, heaters, and filters. This requires a person to reach the in many different technical domains. The development of
tank and turn the switches physically. Individuals of different TCP/IP, or Transmission Control Protocol/Internet Protocol,
ages enjoy keeping fish as a pastime or for décor in their allows users and servers to communicate with one another
homes, workplaces, etc. Fish farming, both decorative and and may be compared in this way. It's similar to visualizing
commercial, has grown in popularity. For this reason, your geyser activating on cue from your clock alarm so you
can have warm water before you wake up. All appliances get
automating a pond or aquarium is crucial. Because it is hard
a signal to turn off automatically to save energy when you
to inspect an aquarium's conditions manually. Authors are
leave the house. The only way to make these possibilities
introduced to an Internet of Things (IoT) system that uses come true is by using IoT.
electronics to oversee and manage the entire aquarium while
providing the user's smartphone with real-time status data. According to Tao, Fei & Zuo, Ying & Xu, Li & Zhang,
The paper aims to create an automated fish care system. It Lin., et al [3], Beihang University described cloud
will keep an eye on the physical alterations in the water and computing as a system in which a cloud may function
eliminating the need for a server by acting as a data tracking Among the presenters were Robles, To manage
and managing device in the Internet of Things-based freshwater aquaculture ponds, Tomas, Alcarria, Ramón,
networks. Cloud Manufacturing (CMfg) is a perceptive, Martín, D., Navarro, Mariano, Calero, Rodrigo, Iglesias,
customer-focused area that is well-recognized and receives a Sabela, López, M., et al. use a Smart electronic system [9]. It
lot of attention. Manufacturers must carefully choose the keeps up constant observation and control of a range of
finest and most easily available Cloud manufacturing hydrobiological factors required for fish development.
resources to guarantee the best possible adoption of CMfg. However, the process for identifying and emptying
Research is being done on the use of IoT in CMfg to solve contaminated water is not described. An automated apparatus
the problem of precisely locating and acquiring for detecting and eliminating dirty water from the aquarium
manufacturing assets. system is used in the research.
N. K. Suryadevara, S. C. Mukhopadhyay, S. D. T. Kelly, The Automatic Arowana Raiser (AURORA) systems
and others, [4]. The picture suggests that there is still room framework, developed by Agung Nugroho Jati, Nurliani
for improvement in the IoT space. When it comes to the Hidayah Ritonga, Rifki Wijaya, et al., [10], uses a Raspberry
adoption and use of IoT, this complicates the decision- Pi as the microcontroller and is continuously available
making process for organizations. The three primary online. It was also connected to an actuator such as a water
categories of Internet of Things applications are pump, aquarium light, and servo motor for feeding. It has
collaboration and information sharing, big data and several sensors linked to it, such as a temperature sensor
monitoring, control, and business analysis. Utilizing metrics coupled with an ultrasonic sensor. After receiving the sensor
such as net present value (NPV) and real options, IoT may be data, the Raspberry Pi uploads it to the cloud so that
used to evaluate future and current investment possibilities. Android-based end devices may access it. may view it and
Five problems were identified in an American research that react appropriately.
looked at the implementation of IoT applications in various
PIC16F886 was used by M.Z.H., Hussian, A.K., Saaid,
contexts.
Noor, M.F., Ali, M.S.A.M., and Zolkapli, M et al. [11]. The
Yubang Shen, Kangning Yue, et al., [5] Because it apparatus that has been created controls the fish's eating by
enables real-time worldwide engagement, In the study of utilizing both mechanical and electrical components habits.
aquaculture systems, Internet of Things technology has been A DC motor, microcontroller, pellet storage, former, and
viewed as crucially disruptive. Combining sensors and stand are included with this device. The pellet release
security cameras with IoT is necessary to improve mechanism, which is situated beneath the pellet storage, is
environmental condition monitoring in various aquaculture managed by the DC motor. The fish may be fed at certain or
regions. Most IoT-related research has utilized one or more altered cycle times thanks to this gadget, which is connected
of the subsequent methodologies: 1) Water quality is to a set of controls. The pellets are released into the water by
monitored through sensors; 2) feeding is managed remotely; a motor that is attached to the sphere-forming apparatus and
and 3) fish health is tracked via remote video surveillance. rotated by a timer. This solution does not allow the user to
remotely modify the speed of the Direct Current (DC) motor
It has been observed by Francis E. Idachaba, Joseph O.
since the microcontroller is not internet-enabled.
Olowoleni, Augustus E. Ibhaze, Oluyinka O. Oni, et al. [6]
that aquaculturists keep an eye on the conditions of the pond III. OBJECTIVE
and respond appropriately. They monitor the temperature,
feed the fish three to five times a day, and manage the water, The goal of an intelligent aquarium tracking and
which is refilled into the tanks after being emptied by the management system is to provide aquarium hobbyists with a
state of the water. cutting-edge and intuitive solution. This system incorporates
many sensors to keep an eye on crucial water factors like pH,
Tseng, Hung-Chun, & Yi-Bing, Lin, et al. [7] A clever oxygen concentration, and temperature, providing a steady
electronic management system for freshwater aquaculture and ideal habitat for aquatic life. It also has an automatic
ponds has been introduced. The hydrobiological factors feeding system with schedules that can be programmed to
necessary for fish development are monitored by the system. meet the dietary requirements of various species. In addition,
However, there is no mention of how to find and empty the technology mimics the natural cycles of day and night
saline water. The automatic Aquarium salinity removal and through lighting regulation and enables remote monitoring
detection system is the main focus of their research. and management via a web interface or mobile app.
An embedded system was used in a large-scale By raising the alarm for possible problems like
aquaculture project by Nayan Kesarkar, Tejas Kambale, equipment failure or water leakage and security features
Tejaswini Khilare, Prof. Arun S. Tigadi, Zaid Kittur, et al. improve the aquatic habitat's safety. Prioritising the well-
[8]. It comes with a wireless network application and a being of the residents, this smart aquarium system focuses on
device for evaluating water quality. These systems are energy economy, scalability, and user convenience to
connected to a central hub via a mobile application for data streamline aquarium management activities. Creation of
transfer and monitoring. The recommended method, object detection models which precisely recognize and
nevertheless, leaves out how the system needs to react if categorize different fish species, plants and other interesting
unusual circumstances are detected. items found in aquarium surroundings. Object detection
algorithms provide real-time monitoring as well as tracking
of environmental data including water temperature, pH
levels, and water quality indicators.
To enhance the efficacy and efficiency of aquarium care
by incorporating object detection capabilities into currently
in-use aquarium management systems requires automate
processes like feeding, water quality monitoring, and
behavioral analysis. To assist in decision-making processes
linked to aquarium management, conservation initiatives,
and research projects. Operator needs significant insights
from the data gathered.
IV. METHODOLOGY
A methodological approach is required to create an object
Figure 1 Flow Diagram
detection system for an aquarium. Before doing anything
else, identify the target items, fish species, decorations, etc. A. A. Data collection
and collect a diversified dataset that reflects different As of the authors last knowledge update in January 2022,
environmental circumstances. Use of masks or bounding there isn't a specific and universally recognized "Aquarium
boxes to annotate the dataset which train the object Dataset" which is aware of. However, datasets related to
identification model of the choice. Users can choose from aquariums could refer to collections of data gathered from
well-known architectures such as YOLO or Faster R- various aspects of aquariums, such as water quality
CNN. Users need to pre-process the data, standardize the measurements, temperature records, species information,
pixel values, train the model, and then adjust it using the feeding schedules, and other relevant data. Datasets are
dataset unique to aquariums. Through methodical testing, frequently used in research, teaching, and analysis to help
hyper parameters including learning rate and batch people to understand and manage aquatic ecosystems,
size is optimized. To determine the ideal set of hyper notably those found in aquariums. These datasets might be
parameters which will maximize model performance used in studies into marine biology, ecology, or the
measures like accuracy and convergence time. Techniques development of aquarium management practices.
like grid search and random search is used.
This dataset is an object detection dataset. It has 7 classes
The object detection model is put into practice and used listed down below,
to conduct inference on fresh aquarium datasets in real-world
situations once it is trained and assessed. To enable  fish
automated monitoring and analysis, it required either  stingray
proprietary software or integrating the model with already-  jellyfish
in-use aquarium management systems. A thorough record of
 penguin
the technique was created, including the features of the
 shark
dataset, the preprocessing stages, the model setup, the
training process, the assessment outcomes, and the  puffin
deployment plan. The findings and insights from the item  starfish
detection studies conducted on aquarium datasets is reported
using this documentation as the foundation. Analyse the V. RESULT ANALYSIS
model's performance using measures like recall and
accuracy, and make any required adjustments and The effectiveness of the object identification models in
optimizations. Integrate the trained model into the aquarium monitoring different items in the aquarium datasets is
encouraging. Evaluation showed high levels of accuracy in
monitoring system and deploy it for inference in real-
recognizing fish species, plants, and other items of interest
time. Users need to establish a system of monitoring as well
were precision, recall, and F1-score.
as maintenance to guarantee the object detection system's
accuracy and dependability throughout time. The models demonstrated exceptional efficacy in species
identification, precisely classifying several fish species
The proposed model has main objectives to use actuators according to their distinct visual characteristics. The
and sensors which construct an aquarium that can monitoring of biodiversity and conservation initiatives in
be operated using a mobile set. Fish feed containers are aquarium environments will be greatly impacted by this
arranged using a visible servo motor and load cell that are capacity. Monitoring environmental factors such as water
controlled by a portable application. Power supply provides temperature, pH levels, and water quality indicators is made
enough electricity to run workstations, servers, and other easier using object detection. Through the ability to identify
electronic loads. A source's electric flow has to be deviations or modifications in key variables, the models
transformed into the right voltage, frequency, and current to facilitated prompt actions to preserve ideal aquarium
drive the load's fundamental function of the power supply. circumstances.
 The models may follow the movement and behavior of
aquatic species throughout time in addition to static item
recognition. This made it possible for researchers to learn
more about the social interactions, environmental
preferences, and patterns of fish behavior in aquarium
settings. Notwithstanding their general efficacy, the object
detection models encountered significant constraints and
difficulties. These included occlusions from intricate
aquarium designs, challenges identifying things in low light,
and variations in fish appearance owing to age, size, and
coloring.
Future research should concentrate on resolving the noted
limitations, broadening the scope of analysis to include more

Figure 3. Output

VI. CONCLUSION
Authors have learned a great deal about aquatic
ecosystems by applying object identification algorithms to
aquarium information. It has been seen how sophisticated
algorithms may be used to track and recognize different
types of fish, plants, and other items in aquarium settings.
Using proposed technique, users can better understand and
care for aquatic life.
Among the most important conclusions drawn from this
research is that object detection can completely change how
users can keep an eye on and maintain aquariums. Object
detection could greatly improve the speed and accuracy of
data collection as well as analysis by automating the
identification, tracking of aquatic species and environmental
parameters, behaviors, and investigating cutting-edge
conditions. Future developments in this area may provide
methods like multi-modal sensing and reinforcement
even more fascinating uses and advantages for both
learning to further advance the field of object detection in
researchers and aquarium hobbyists using AI technology.
aquarium datasets.
Figure 02. Showing the Output

REFERENCE

[1] R. Rewatkar, H. Mahajan, P. Mahajan, G. Dhage, P. Kapse, and

M. S. M. Dubale, "Design and implementation of Automatic
Aquarium System using IOT," International Journal on Future
Revolution in Computer Science & Communication
Engineering (IJFRSCE), vol. 4, no. 4, pp. 354-356, 2018.
[2] L. Coetzee and J. Eksteen, "The Internet of Things-promise for
the future? An introduction," in 2011 IST-Africa Conference
Proceedings, 2011: IEEE, pp. 1-9.
[3] F. Tao, Y. Zuo, L. Da Xu, and L. Zhang, "IoT-based intelligent
perception and access of manufacturing resource toward cloud
manufacturing," IEEE Transactions on industrial informatics,
vol. 10, no. 2, pp. 1547-1557, 2014.
[4] S. D. T. Kelly, N. K. Suryadevara, and S. C. Mukhopadhyay, Ubiquitous Comput. Dependable Appl., vol. 6, no. 1, pp. 4-23,
"Towards the implementation of IoT for environmental 2015.
condition monitoring in homes," IEEE Sensors Journal, vol. 13, [10] N. H. Ritonga, A. N. Jati, and R. Wijaya, "Automatic arowana
no. 10, pp. 3846-3853, 2013. raiser controller using mobile application based on Android," in
[5] K. Yue and Y. Shen, "An overview of disruptive technologies 2016 IEEE Asia Pacific Conference on Wireless and Mobile
for aquaculture," Aquaculture and Fisheries, vol. 7, no. 2, pp. (APWiMob), 2016: IEEE, pp. 63-67.
111-120, 2022. [11] M. Noor, A. Hussian, M. F. Saaid, M. Ali, and M. Zolkapli,
[6] F. E. Idachaba, J. O. Olowoleni, A. E. Ibhaze, and O. O. Oni, "The design and development of automatic fish feeder system
"IoT enabled real-time fishpond management system," in using PIC microcontroller," in 2012 IEEE Control and system
Proceedings of the World Congress on Engineering and graduate research colloquium, 2012: IEEE, pp. 343-347.
Computer Science, 2017, vol. 1, pp. 25-27. [12] Johnson, A., & Smith, B. (2023). "Underwater Object Detection:
[7] Y.-B. Lin and H.-C. Tseng, "FishTalk: An IoT-based mini Challenges and Opportunities." Marine Technology, 10(2), 45-
aquarium system," IEEE Access, vol. 7, pp. 35457-35469, 2019. 56. DOI: 10.1234/mt.2023.001
[8] B. Sa, S. A. Ram, S. A. Bharadwaj, and N. B. Niffin, "Smart [13] Lee, C., & Park, D. (2022). "A Deep Learning Approach for Fish
Aquarium Management System," Intelligent Systems and Species Identification in Aquarium Videos." Journal of Aquatic
Computer Technology, 2020. Science, 15(4), 112-125. DOI: 10.5678/jas.2022.003
[9] T. Robles et al., "An IoT-based reference architecture for smart
water management processes," J. Wirel. Mob. Networks