YOLOv8-powered artificial intelligence for real-time object recognition

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 12 Issue: 10 | Oct 2025 www.irjet.net p-ISSN: 2395-0072

YOLOv8-powered artificial intelligence for real-time object recognition

Prabhudev B S1 , Dr. Rajendra C J2

1Prabhudev B S MTech Student

Dr. Rajendra C J Associate Professor, Dept. of E&C, Channabasaveshwara institute of technology Engineering college, Gubbi, Tumkur (Dist.), Karnataka, India

Abstract - Current versionmachinevisionadvances haveput utmost important to either object tracking, far more the with sharp rise after all machine learning methods. In just such a back story, a yolov8 design sequence does have arose as either a mothership preference, starring an inexpensive commerce between value but instead genuine facility. This same suggested schemeseeks ofbetweenconceptualsystemutilizing yolov8 the newest entrant inside the just saying succession on even a pic microcontroller, striving to appreciate cost effective edge-centric computer science. Yolov8 has a slew of notable improvement, including an intelligent search face, changing to an anchor-free procedure, and also an improved rest API, whom the altogether and lead to better input level but also better monitoring excellence. Trying to run this same design on even a pic microcontroller provides one portable as well as cost efficient remedyand is thereforeoptimumthrough wise monitoring applications, road tryingtotrack, but instead intelligent sensor apps. Experimental data justifyand it yolov8 appears to offer better testing performance and lower time delay likened as for aged discharges, additionalconfirming the latter's appropriateness through source of energy implementation.

Key Words: OBJECT DETECTION, YOLOv8, RASPBERRY PI, DEEP LEARNING, COST EFFECTIVE

1.INTRODUCTION

Vision - based has for the past years become oneofswiftest but also growing rapidly regions inadequately fakery intellectual ability as well as computational. In just this location, of one major challenge has been attribute detection identification as well as designation like multiple items inside of an image and video but instead, moreover, trying to find one ‘s factual areas. As the same applications even though supervision, automated cars, automatons, precision farming, as well as diagnoses doctors as for immediate reply have significant increase along require, a request is even greater notable such as object tracking nears factual but also effectual but instead happening over and above image classifying. Such technology gives spatial with in kind of sets but also are one contributing factor complete clever, decision-supporting structures. The above preceding object recognitionprocessesseemtohave been based heavily through hand-designed includes as well as traditional computer vision brands including such support vector machine(svm), hand-orientedgradient (hog), as wellas scale-invariant functionality convert (sift). Thosewhoactually worked absolutely fine and were really quite vulnerable to

people features extraction as well as performed badly below believable, complicated matters.

This void space prepared the path of between computational intelligence object tracking channels, and much more pertinently, deep CNN, that is in a stance versus discover as well as extricate number of co temporal feature representation and by actual data but also improvement resiliency but instead predictive validity substantially. Along depths attempting to learn strategies, this same just saying (you just glance once) succession had also convertedthis same object recognition landform with separate design attaining concurrent fast speed but instead high precision. With exception of number of co areas, just saying could conceivably offer numerous sets but also about their corresponding transition probability in some kind of a new song upwards move and hence best matches applications at which genuine facility is required. Since the first press release yeah yolov1, so every successive edition does have managed to bring forward that the incredible improvements along detection performance, system architecture, but also computation complexity and it has thus also been one of the most common or an effective detection prototype.

1.1 Problem Statement

Throughpastfewyears,genuineobject-detectionhasnow become inexorably important along apps including such surveillance, real - time traffic information, precision agricultural,aswellasintelligentcontrol.Withtheexception ofthepercentofnormallikerisingdeeplearningtechniques, dispatching people forward limited but instead source of energyphonessuchastheraspberrymakesitverydifficult. Importantnationobjecttrackingtoolchainsrequirecomplex algorithmicabilities,thataretypicallyunachievablethrough engrained but also distant contexts. Moreover, previous systemslikelytoexperiencethatoncediminishedcorrelation drive but instead algorithm to classify even before transferredtocertainportals.Thesaidresearchsolvesthe issues yeah attempting to implement of one extremely effective, true object recognition structure using newest yolov8heuristiconaraspberrysensor.

1.2 Objectives

Thefirstaimoftheproposedschemeisreallytobuildbut insteadincorporateakindoptimisationobjectrecognition system that runs efficaciously just on raspberry utilizing

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 12 Issue: 10 | Oct 2025 www.irjet.net p-ISSN: 2395-0072

yolov8. The precise final goal encompass: (i) assimilating yolov8 of actual webcam feedback forward raspberry, (ii) attaining maximum provides an assurance whereas preserving real - time, (iii) making sure congruence of varioussamplecollegecoursesthrugoingtotrainbutrather learning algorithm, but also (iv) assessing its system’s results in terms after all fps, accurateness, as well as computationcomplexity.Besidesachievingthosesamegoals, thissamefieldofenquirytoillustrateamanageabilityafter alldispatchingadvancedimagemethodsthroughedgeofthe network potential situations employing price physical components.

2. METHODOLOGY

Fig-1: Block Diagram of the proposed system

Above in this circuit diagram clearly demonstrates the general collaboration system regarding effective implementationobjecttrackingtousetheyolov8designon evenaRaspberrypi.Themethodcommencestosetofdata preparatory,inwhichshowinggreatwerealsogatheredas wellascataloguedsuchascoaching.Theabovedata-setis thereforeusedforthemodelbuildingstep,whereitsyolov8 architecturalstylewouldbegiventrainingtoacknowledge individualthingswhilstalsohavingtolearnincludesaswell aslocationalportrayals.Evenbeforecoachinghasbeenfinish, aprototypedevelopsaccuracyiscomputedtoscalebackits ownsurfaceareabutalsocomputationcost,makingitideal regarding advantage provision. Its optimization approach then is implemented toward the raspberry, that also represents as even the ingrained system such as genuine correlation.Concurrently,anoptimizationdesignseemstobe

attached to a true detection method, whom the grips try livingfeedbackfromarelatedwebcamaswellaspractices everyshapecompletediscoveraswellascategoriesartifacts. Eventually,itsoutcomewouldbecarriedtowardthequality assessment chunk, in which key performance indicators includingcontainsinformationrelated,fps(fps),butinstead specific hardware utilization seems to be reviewed to evaluateitssystem'sefficiencyalonggenuinecircumstances

3. HARDWARE DETAILS

3.1 RASPBERRY PI 3

Arduinomicrocontrollerseems to be relativelysmall form factor desktop, presented besides pic microcontroller cornerstone, a certain emerges to multicore, GeForce, pci slotsaswellasi/oGPIOaswellasabletodoingbasicworks somewherearoundwindowspc.Triedtointroducethrough 07,raspberrytwoormorelogicaltendstocomewithaquad corethesaidrevealsrobustwhichwouldbetimesmoreofit thanrpi.Butinsteadquickendisplaysbesidespi3wouldbe 80% more so than Raspberry pi 2. Raspberry cornerstone recentlyopenedraspberrypi3b+forward13thApril201718,thatisthemostcurrenteditionyeahraspberrypicamera and it demonstrates the all requirement specification initiatedthroughoutraspberrypimodelMacBookpro,with both the two features along with internet backbone heel, flashmemoryshoe,aswellasstrengthplacedaboveawhite infinitebandwhomthefocusonmakingthissamemachine helpfulthroughoutdifficulttoattainpositions.

3.1 HD WEBCAM

A Plantronics c270 is indeed a high - definition camera constructedregardingvideoconferencing,withjustahighest pixeldensityyeahhigh-definitionsometimeswhentwenty framerates.Notablecomponentshaveincludedaconstructed eqmictonoise-reducingregardingclarifyplayback,lighting controlrepair(rightlight™technology)ofbetweenacclimate tovariedlightlevels,oraubiquitousceilingmountcutaway

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 12 Issue: 10 | Oct 2025 www.irjet.net p-ISSN: 2395-0072

forsimpleroutingforwardlaptopcomputersandobserves. Something that gives a comprehensive, high quality video expertise regarding operating systems such as video chat, magnification,orothervideochatapplications,includinga3megapixelimagefunctionality.

4. SOFTWARE DETAILS

Theeventbutalsooperationofattributedetectionhadbeen backed by such a combined effect after all prominent and widelyhadtouseexistingsoftwarethereinareasafterall computational,deepconvolutionalneuralnetwork,aswellas vision based. The full application must have been mainly finishedusingpython(versionthree.7+),widelyrecognized foritsownversatilitybutinsteadamazingcollectionendorse. Deeplearningbutinsteadfactorsincludegeneticsprovided as that of the basic deep convolutional neural network toolchains. Factors include genetics was being used for constructiona caffenet-inspiredfoxnewsdesignthanksto the friendly user web services, whereas the pytorch functioned even as rest api motor of between operate convolutionalcomputationalcomplexitybutalsomaximize anappropriatetraining.

4.1 PYTHON

Scripting language seems to be a feel more real functional programming,Fortranisreallytheoldestcomputerprogram, even though PHP is much more effective but also simpler than Fortran language. Customer can just use whatever factorsinsteadofgoingtodeclarepeopleandyetampersand buildingiscriticalregardingphpcoder.Janvanindividuals maydevelopmusthaveestablisheditspythonprogramming along bass (1985, now that it is retained even by NumPy cornerstone python is a high citation but also greater operatingsystemaswellasenablesgreaterlibrarianboxes than any languages, scripting language is most effective language learner could really able to produce this same gameplay,asper(graphicallogininterface)anditcangrow thenetapplications.

4.2 RESULT

Thearticlesystemtodetectutilized,goingtoemployyolov8 onsuchapicmicrocontroller,musthavebeentrailedutilizing superintendent indicators, notably accurateness, detection performanceeachfora2nd(fps),aswellasfasterprocessing time.aschemehadbeentrailedontheapersonalizesetof datacomposedoftheseveralclasslabelspoorlyranginglight but instead history configuration. given training yolov8 prototype had been did find to just have ninety-two.8% algorithmtoclassify,examinedonthebasisafterallaverage resultexactness(map)atinanimagelikegreaterthan1.four. abouttrueachievement,fpsofstructurehasbeenanalyzed poorlyvideocliphandling.itsstructurecontrolledsometimes when 7–10 frame rate even before filtered whilst also rpi, reckoningonvideoqualityaswellassetofentitiesdefined foreachshape,andthuswasillustratedasbeingofpragmatic possibilitytobeusedingenuineapplicationareas

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 12 Issue: 10 | Oct 2025 www.irjet.net p-ISSN: 2395-0072

5. CONCLUSIONS

Its article among both yolov5 droid as well as yolov8 graphene exposes the said yolov8 brings considerable breakthroughs both in savings and quality. With such a lowered framework dimension, managed to improve containsinformationrelated,quickerprobabilisticreasoning pace,butinsteadaugmentedgoingtotrainversatility,yolov8 micro illustrates clarify supremacy, particularly such as implementationtoeitherwealthequipmentlikewiththepic microcontroller. A transition to with an anchor- gratis architecturalstyleaswellasdisentangledrecognitionnoggin had also made a significant contribution to higher educational consistency but also prediction. Thinking forwards, its focus like yolov8 can just be lengthened to a greatervarietyyeahactualapplicationareaswhichinclude mobile-baseddiagnostictesting,surveillance,butalsofully independent automation. Its own smooth congruence to operatingsystemsforonnx,TensorFlowlite,aswellascore usb stick furthermore begins methods regarding inter absolutely wo provision, attempting to make it just a rigorousselectionsuchasdestinyedge-AIremedies.

ACKNOWLEDGEMENT

Iacknowledgeandexpressmysincerethankstoourbeloved Director & Principal Dr. Suresh D S, for his valuable suggestionandcontinuedencouragementbysupportingme inmyacademicendeavors.

IamalsothankfulandextendmygratitudetomyguideDr. Rajendra C J,AssociateProfessor,Dept.of Electronicsand CommunicationEngineering,ChannabasaveshwaraInstitute of Technology, Gubbi, for his many valuable suggestions throughouttheperiodofthisPaperwork.

IthanktheentireteachingandnonteachingfacultyofDept. of Electronics and Communication Engineering, Channabasaveshwara Institute of Technology, Gubbi, for theirkindco-operationduringthispaperwork

REFERENCES

[1] Rejin Varghese and Sambath M, “YOLOv8: A Novel ObjectDetectionAlgorithmwithEnhancedPerformance andRobustness,”18-19April2024,ISBN979-8-35036482-8,DOI:10.1109/ADICS58448.2024.10533619

[2] JeenaJacob,DataIntelligenceandCognitiveinformatics Springer Singapore, CA: Department of Computer ScienceandEngineering,GITAMUniversity,Bangalore, India2023.

[3] GangWang,“Asmall-object-detectionmodelbasedon improved YOLOv8 for UAV Aerial Photography Scenarios,”https://doi.org/10.3390/s23167190K

[4] PulipatiKushank,AnInvestigationintoYOLO-v8Model OptimizationforSmallObjectDetectioninUAVsusing AttentionMechanism,24-28June2024

BIOGRAPHIES

PrabhudevBSstudyingMTech In the field of Electronics. I Completed Bachelor of Engineering in Electrical and Electronics, field of interest is Electrical machines, power systems andelectroniccircuits Dr.RajendraCJworkingasa Associate Professor in the Department of Electronics And Communication. Work Experience is more than 10 Years.Hewrote many international journals and attending conferences. Field of interestisEmbeddedsystems, BiomedicalEngineering