Virtual Dressing Room with Web Deployment
Dr. S.Palanivel Rajan1, V.Hariprasad2, N.Purusothaman3, T.Tamilmaran4 Associate Professor1, UG Student2,3,4
Department of Electronics and Communication Engineering, M.Kumarasamy College of Engineering, Karur, Tamilnadu, India. Email ID : drspalanivelrajan@gmail.com
Article History: Received: 11 January 2021; Revised: 12 February 2021; Accepted: 27 March 2021; Published online: 16 April 2021
Abstract: Trying different clothes in shops and finally selecting the right one is a time consuming andtedious task. So, a Real-time virtual dressing room is the concept where the customers can buy clotheswithoutwearingthem.Virtual dressingenvironment is the online equivalent of the in-store changingroom. People usually avoid buying wearable clothes online mainly because it's hard to judge
whether itwilllookgood on them or not. To solvethis problem,wedecidedtobuildanOnline
TrialRoomApplication. Our research is focused on developing an application that uses the system camera to capturea video of the user and then splits the video into individual frames from which the user's body is extracted.Finally using functions to extract information on the placement of joints in the body and to transform,rotate, and scale the wearable image onto the user in real-time. In the literature review, we go throughvarious ways to achieve our goal with their advantages and disadvantages. The project is implemented inFlask Web application with OpenCV a Python Module. The application works on devices with an inbuiltor attached camera, internet, and web browser. This project introduces Augmented Reality-based VirtualTrial Room software that allows users to digitally wear clothes by superimposing 2D clothes over devicesand virtual clothes over the monitored user. The clothing moves and enlarges in response to the user'smovements.2DAugmentedReality is used in this web implementation.
Keywords: OpenCV, VirtualdressingApplication,2DAugmentedReality
1.INTRODUCTION
Purchasing wearable’s online is always a risky endeavor because one never knows how the item will lookon oneself. Furthermore, purchasing clothing or ornaments from a store that does not online takes a longtimebecauseyoumustfirstlocateastore,thengointothetrialroomandtryoneverycloth. By digitizing the process, the proposed solution will help users save time while testing out wearables. Itwas chosen to use OpenCV because it is much easier and has been pre-trained to detect the user's bodyover which the fabric will be superimposed, saving them time while offering an excellent user experience.The user will receive results in real-time, i.e., the output of the wearable superimposed will be givensimultaneously with the input by capturing every frame of the video and applying the attire to the user'sbody in that video frame, then returning the frame to the user, giving the impression that the results aredisplayed in real-time. Unlike some of the proposed works in section II, the implementation does not needany hardware, making it a very cost-effective solution. The proposed application is also platform agnostic,meaning it can run on any operating system on any computer that has a webcam, internet connection, and web browser. The current project is focused on the "VIRTUAL TRIAL Space" virtual reality program. The virtual trialroom is software that allows the user to put on and see the dress from a distance by simply standing in front of the sensor. People nowadays go to shopping malls and textile stores to put on dresses in a testing room to see if the dress suits them perfectly before purchasing it. As a result, people must go to the courthouse and wait for the argument to be heard to become available so that the user can enter and look at his clothing. Since thenumber of rooms of trials available in most textile stores or shopping malls is small, it is extremelycrowded. Often, people must spend a significant amount of time waiting for the trial space, which is awaste of time.Trial roomsaren'tcompletely safe,so no one can be certainthere aren'tany hiddencameras.As aresult,changingclothesinthecourtroomposes adirectthreattowomen's safety.
In the Virtual trial room app, the user can change his clothes virtually using virtual clothes. To use thecamera, the user must stand in front of it. Using algorithms, the camera scans the human body from theenvironment. The consumer is shown on the monitor that is linked to the Web camera and the deviceprocessor, as well as a list of virtual dresses. The Web camera scans the scene and displays live videostreaming in the window. A list of dresses that the user should wear is also shown on the monitor. Whenthe userchoosesadress,the dressischosen and the userwearsitvirtually.By scanning the user'sskeleton joints, the dress is superimposed over them. The tailored dres' moves in sync with the user'smovementsinfrontof theWebcamera.
As a result, this virtual trial room tech may have a major impact on today's shopping experience. Peopledo not need to queue for a long time in front of the trial room to check out their dresses, nor do they needto be concerned about hidden cameras. And it allows people to change their clothing or try on dresses in afraction ofasecond.Mostoftheuser's timeissaved,andtheireffortisgreatlyreduced.
Theproject'stwomainissuesare:
1. Wearable’s superimposing accuracy by the consumer.
2. Arealisticperspective.
1.1 SUPERIMPOSITIONACCURACY
The first concern of a virtual trial room is crucial because it determines the accuracy of the entire system.The accuracy is highly dependent on the application's algorithm's ability to locate the user in the videoframes. The first method is to use neural networks to train the algorithm to find the human body in theframe, and the second method is to use a marker such as an RGB color to pinpoint the user in the frameusing color pixels. The latter is less user-friendly, but OpenCV already has a trained algorithm forrecognizingbodypartssuchastheface,upperbody,andlowerbody.
1.2 VIRTUALTRIALROOMWITHAREALISTICVIEW
Computer-generated Reality execution is pointless on the off chance that it doesn't feel regular, which isjustconceivableifthebuyercanencounterasimilarsensationastheydowhilewearingatexture,whichis not the same as the vibe of wearing a cotton material against a woolen fabric. Regardless of whether itcan't give that level of authenticity at this moment, it can in any event make the client's view moretrustworthy,asthoughthey'retakingastabatthematerialinamirrorinsideagenuineTryonspace.
The expression "increased reality" alludes to an immediate oraberrantperspective on truecomponentsthat have been upgraded utilizing PC programming. Valid and mimicked segments are consolidated inAugmentedReality.Itfundamentallyconsolidatesprogramminginformationandrefinestheclient's perspective on the genuine world buyer can see both virtual and characteristic light in most enlargedreality games. This is refined by utilizing extended pictures to layer pictures and intuitive virtual items ontopoftheclient'sperspectiveonthispresentreality.IncreasedRealityframeworksarefrequentlyindependent, untethered, and needn't bother with a link or a PC to work. OpenCV is an abbreviation forOpen Source Computer Vision Library, which has interfaces in Python, C++, and Java. It is principallyexpected to improve computational execution while likewise stressing constant applications. At the pointwhen the code is written in C or C++, this bundle has the extra advantage of multi-center handling.Clients' time is saved and the disarray caused during the acquisition of wearables is decreased by utilizingenlargedrealityinnovationtocarefullyputthemon.
2.
LITERATUREREVIEWMuhammad Kotan introduced a paper about the picture handling and virtual changing area application"Virtual Mirror with Virtual Human Using Kinect Sensor" permits clients to attempt virtual pieces ofclothing before a virtual mirror. In a virtual dressing space, a virtual rendition of the dress shows up. Thegarments are chosen from a rundown on the screen by the client's hand movements. From that pointonward, in the virtual mirror, the picked virtual garments show up on a humanoid model. The Kinect isutilized to check the components of the client. The 3D areas of the joints are utilized for situating, scaling,and pivoting to adjust the 3D articlesof clothing to the model. We would now be able to show andvitalize virtual people on account of PC designs. Continuous perception and movement are needed tomimicpeople,consideringinformationimperativesforthesevirtualpeopleaddressingclients.Bychecking the client with the Kinect, a 3D model of a person is made. The client is examined utilizingAGPL3.0 programming to make a 3D model of the client, a humanoid model, and solidarity to create 3Dmodels of pieces of clothing. The console and mouse are utilized to connect with the framework by theclient. Since the application is about a 3D model of the client, it doesn't address the issues of the client.The client can't partake in the live dressing idea. These thoughts won't assist with web-based shopping.Just the application can utilize the piecesof clothing planned in solidarity. Clients can take a stab atpieces of clothing on a made humanoid
model because of the changing room's GUI (Graphical
UserInterface)perusinganddecipheringinformationfromconsole,mouse,webcam,orKinect inputunits.
3.
EXISTINGSYSTEMCustomers usually have to travel to tailor shops to have their measurements taken for garment tailoring.Their information is recorded and stored on paper. Customers must also leave their offices and inspect theclothing to see whether it is full or not. This is both time-consuming and expensive. The whole process issluggish due
can openly investigate all garments in advanced photos or 3D models in onlineshops, yet they can't encounter the impacts of getting into garments. Clients can take a stab at any imageofgarmentsin2Dvirtualdressing,yettheywon'tfeel3Dfittingbecauseofcomputerizedpicturesfrom
clients and pieces of clothing.Since both the customers and the clothes are 3D models, 3D virtualdressing allows for 3D fitting. Customers can freely try on any clothing model in 3D spaces, as well aschoosetheirdesiredscale,color,andtexture.
4.
PROPOSEDSYSTEMBy separating the userregion,the usercan create an atmosphere and context from the video stream,which can then be virtually layered onto the user interface. It is also useful for skin identification and fordetermining the area of interest.The human body is addressed by body part joints like the head, neck,hair, and arms. A considerable lot of the body parts' joints were addressed in 3D directions. Thus,
thesectionalistreatedasacloseoutstrategyfortheprofunditypicturedependentontheper-pixelcharacterizationtask.Acombinativejourneycanbeevadedifeverypixelis isolatedindependently.
Python Flask Web Application Interface was used to build the application. On the website, the user canlook at clothes and other wearables and decide whether to buy or put them on. The user must press the‘Quick View' button if they want to try on the wearables online. The Tryon script will be executed as aresult of this. Video is shot utilizing the framework camera, and the clothing picture is superimposed onthe client's body continuously utilizing OpenCV. If the client likes the outfit, they can purchase it or shopotherwearablesonthewebsite,muchastheywouldinaphysicalstore.
FIG1:ProposedBlockDiagram
4.2 TESTING ANDMAINTENANCEOFTHESYSTEM
Various forms of testing are performed during the production process. Each test form is designed toaddressaparticular testingneed. ThefollowingshouldbeincludedinaproperAndroidresearchstrategy: 1. Unit Test 2. IntegrationTest 3. FunctionalTest 4. SystemTest 4.2.1 UnitTesting:
The unittestisthe firsttestin the production phase.The source code isusually brokendown intomodules, which are then broken down further into smaller units known as units. These units have distinctpersonalities. Unit testing is a type of test thatis performed on these code units. The language in whichthe project is written influences unit testing. Unit evaluations ensure that each project's distinct coursefollows the documented criteria and has specified inputs and planned outcomes. The Junit 3.0 system ispre-installedontheAndroidplatform.It'sanopen-sourceplatformforunittestingautomation.
4.2.2 IntegrationTesting:
Modules are integrated and tested as a group during training. Code modules, individual programs, sourcecode, and other types of modules are common. On a network, destination applications, and so on. Devicetesting occurs after unit testing and before integration testing. After the product has been coded, it is timeto testit. Betas are frequently coursed broadly oreven to the overall population in the expectation thatthey will buy the result when it is delivered. A large number of the modules that have been unit tried havebeen consolidated and approved. Mix checks in Android frequently incorporate checking for similaritywithAndroidsegments,forexample,Servicetesting,Activitytesting,Content Providertesting,etc. 4.2.3 FunctionalEvaluation:
Testing at least two modules along with the point of distinguishing abandons, exhibiting that deformitiesare absent, watching that the module plays out its expected capacities as determined in the particular, andkeepingupbelievethataprogramdoeswhatitshoulddoareonthewholeinstancesofusertesting.
4.2.4 SystemTesting:
A project is made up of multiple modules. If the project is long-term, the modules are written by manydevelopers. Several errors can occur until all of the modules have been integrated. System testing is theterm for the testing performed at this point. Device testing ensures that the whole integrated softwaresystem complies with the specifications. It checks a configuration to ensure that the effects are known andpredictable.
Strategy portrayals and streams are utilized in framework testing, with an emphasis on pre-driven cycleassociations and coordination focuses. A specific equipment/programming establishment is being tried.This is typically done on a COTS (business off the rack) gadget or some other framework that comprisesuniquepartsandrequirescustomarrangementsandadditionallyexceptionalestablishments.
5.1IMPLEMENTATION
5.4.1 RecognizingandSizingtheBody
The initial phase in the proposed Online Virtual Trial Room strategy is to procure reference focuses bygetting the state of the body, head, or neck, contingent upon the wearable. The reference focuses are
Motionrecognitionorskeletondiscovery,whichincludedbreaking downafewcasings forsomedevelopment. The discoveries, in any case, were mistaken and deficient for acquiring reference focuses for review thewearable. Therefore, another recognition strategy was created, which included finding the client's face,changing a reference point at their neck, and showing the wearable fixated on that point. Moreover, anAugmentedReality(AR)markermightbeutilizedtoacquireanotherperspective.Eventhoughthiswas
adequate for little frills like glasses or decorations, it was inadequate to plan the garments to the client'sbody.
Freeman’sCodification
It utilizes a comparative computerized body highlight extraction strategy as seen in to get the client'sstature. The thought is to place the shopper before the camera and keep him at a foreordained distancefrom the outset. Focuses on the shoulders and midsection are removed by the calculation. The size of theclientcan be controlled by estimating the distance between these focusesandknowing the separationfrom the client to the camera. At the point when the image (video outline) is caught, a shrewd edgediscovery channel is utilized to extricate just the body's outline. Since natural information is especiallyhelpless to clamor, Canny edge discovery utilizes a channel where the crude picture is convolved with aGaussian channel. Four channels are applied after convolution to distinguish even, vertical, and corner tocorner edges in the prepared picture. To accomplish a shut outline, morphological capacities are regularlyutilized. At long last, every pixel is
allocated a bearing utilizing an 8-point Freeman chain code,
asdemonstratedinFigure1.Ithasthechoiceofutilizingan8or4chaincode,inwhichcasetheaccompanyingequationca nbeutilized:
Theseriescorrespondingtorows 1-8intheprecedingtableisz= 11,7,6,5,9,13,14,15,whichequals
z 4*(deltax 2) (deltay 2) --- >(1.1)
These qualities can be utilized as records in the table, accelerating the chain code calculation. Everydistinction in the course between continuous numbers addresses a 45o fluctuation, so if the distinction inheading between back to back focuses is determined and more noteworthy than two (90o), a componentpointisdistinguishedandsetapartintheimage.
ek |dj 1 dj| 2 --- >(1.2)
ThisisequivalenttoEq.(1.2)expressingthatthesupreme contrastbetweentwofocusesismorenoteworthy than 2. At long last, the size is determined by estimating the distance between them in theimageandcontrastingitwiththedistancebetweenthe customerandthecamera.
5.4.2 Facial Recognition
At the point when a cliententersthescreen,thefaceistheparticularconstructionthatshouldbe distinguished to recognize the client. Therefore, Haar includes based course classifiers are utilized torecognize the face. Rather than utilizing pixel force esteems, the haar classifier utilizes the distinctionconversely values between adjoining gatherings of pixels. The fluctuation contrast between the pixelbunches is then used to figure the picture's general light and dull regions. It's a strategy dependent on AI.As a result, the course work is gained from an enormous number of negative and positive pictures to fitwell with the calculation. The classifier is shown countless negative (pictures without countenances) andpositive (pictures with faces) to prepare it to separate highlights from them. The advantage of utilizingOpenCV is that it accompanies pre-prepared classifiers for the face, eyes, and grin, in addition to otherthings. It accompanies an indicator and an instructor, permitting us to effortlessly prepare it with ourclassifier for any article discovery. If it finds a match, it returns Rect (x, y, w, h), which means left, top,base,andrightorganizes.
5.4.3 Imagemasking
A portion of the pixel force esteems in the concealed picture have been set to nothing. The pixel force ofthe subsequent veiled picture will be set to the foundation esteem, which is typically zero, any place thepixel power esteem is zero in the picture. Or then again The ROIs for each cut is utilized to depict thecover.
Veiling can be overseen cut by the cut in the ROI toolbox if fundamental. Covering activities in theROItoolcompartmenthavenoimpactonacutwithoutROI.
5.4.4. Identificationofthe edge
There are many methods for detecting edges. It used the Canny Edge detection technique [9] for bodydetection, as previously mentioned. Gaussian filters are used to perform this edge detection technique.These filters eliminate noise from digital images to prevent the processor from making a mistake. This isliable for smoothing and diminishing the impact of commotion on the picture altogether for the processortorunappropriately.
The picture's solidarity angles are not uncovered along these lines.Since picture edges can point invarious ways, including level, vertical, and inclining, this calculation utilizes four channels to identify awide range of edges in an obscured picture. To thin the edge, non-maximum suppression is applied afterthis step. In comparison to actual real edges, this suppression produces very accurate edge pixels. Also,certainpixelscanbeaffectedbynoise,inwhichcaseadoublethresholdisappliedtothem.
5.4.5. AttireScaling
Scaling is the way toward resizing an image to suit the circumstance. As the client moves before thescreen, the clothing ought to change in measure and be set on the body fittingly. At the point when theclient draws nearer to the screen, the image size ought to adjust to the client's decision, however, the realelementsoftheoutfitoughtnottochange.Ontheoffchancethatanindividualisgettingintogarments
with an estimation of size S, the size doesn't conform to a size M or L when the individual moves beforethe mirror. Rising or diminishing the general perspective on the texture is all that requires to be finished.Thescalingapproachisutilizedtoaccomplishthis.
6. CONCLUSION
Finally,a Virtual Trial Room wassuccessfully implemented in Python OpenCV.Thisapplication willhelp users save time by allowing them to try on clothes without having to go to the store. The applicationwill monitor the user's movement and angles about the screen to correctly superimpose the attire onto theuser without forcing the user to align to the device screen, resulting in a better user experience. Onlineretailers and sellers will use the app to market their wearable goods, which would certainly draw morebuyers. Last but not least, there is room for improvement in the application's accuracy, especially when itcomes to fabric, which can be accomplished by snapping several snaps of the cloth at various angles andthen aligning the particular angle of the cloth with the particular angle at which the user is tilted. Gettingclothes indifferentsizes,suchasS,M,andL canincreasethequalityoftheapplicationevenmore.
REFERENCES
1. Nikkisingh,sagarmurad,premlone,vikasmulaje"virtualtrialroom"vishwakarmajournalofengineering research,Volume1Issue 4,December2017
2. S.Palanivel Rajan, “Recognition of Cardiovascular Diseases through Retinal Images Using Optical Cup to Optic Disc Ratio”, Pattern Recognition and Image Analysis Journal, E-ISSN No.: 1555-6212, P-ISSN No.: 1054-6618, Vol. No.: 30, Issue : 2, pp. 254–263, 2020.
3. S.Palanivel Rajan, V.S. Sivaanika, “Extraction of Lung in Region of Interest Using Image Data Interpretation”, International Journal of Advanced Science and Technology, P-ISSN: 2005-4238, E-ISSN: 2207-6360, Vol. No.: 29, Issue No. 4s, pp. 2191 - 2201, 2020.
4. Shreyakamani,neelvasa,kritisrivastava,"Virtualtrialroomusingaugmentedreality",International Journal of Advanced Computer Technology (IJACT), Vol. 3/6, pp. 98-102, Dec. 2014
5. S.Palanivel Rajan, L.Kavitha, “Automated retinal imaging system for detecting cardiac abnormalities using cup to disc ratio”, Indian Journal of Public Health Research & Development, P-ISSN: 0976-0245, E-ISSN: 0976-5506, Vol.: 10, Issue : 2, pp.1019-1024, DOI : 10.5958/0976-5506.2019.00430.3, 2019.
6. T.Abirami, S.Palanivel Rajan, “Cataloguing and Diagnosis of WBC’S in Microscopic Blood SMEAR”, International Journal of Advanced Science and Technology, P-ISSN: 2005-4238, E-ISSN: 2207-6360, Vol. 28, Issue No. 17, pp. 69-76, 2019.
7. S.Palanivel Rajan, C.Vivek, “Performance Analysis of Human Brain Stroke Detection System Using Ultra Wide Band Pentagon Antenna”, Sylwan Journal, ISSN: 0039-7660, Vol. : 164, Issue : 1, pp. 333–339, 2020.
8. C.Vivek, S.Palanivel Rajan, "Z-TCAM : An Efficient Memory Architecture Based TCAM", Asian Journal of Information Technology, ISSN : 1682-3915, Vol. : 15, Issue : 3, pp. 448-454,
Tele-Health Care System for Remote Patients”, European Journal of Scientific Research, ISSN No.: 1450-216X/1450-202X, Vol. No. 70, Issue 1, pp. 148-158, 2012.
10. Y. Lin, Mao-Jiun and J. Wang, “Automated body feature extraction from 2D images”, expertsystemswithapplications,vol.38,no.3,pp.2585-2591,2011.
11. S.Palanivel Rajan, M.Paranthaman, “Characterization of Compact and Efficient Patch Antenna with single inset feeding technique for Wireless Applications”, Journal of Applied Research and Technology, ISSN: 1665–6423, Vol. 17, Issue 4, pp. 297-301, 2019.
12. Vipin Paul, Sanju Abel J., Sudharsan S., Praveen M"VIRTUAL TRIAL ROOM",southasianjournalofengineeringandtechnologyvol.3,No.5,pp87–96,2017
13. M.Paranthaman, S.Palanivel Rajan, “Design of H Shaped Patch Antenna for Biomedical Devices”, International Journal of Recent Technology and Engineering, ISSN: 2277-3878, Vol.: 7, Issue:6S4, pp. 540-542, 2019.
14. S.Palanivel Rajan, R.Sukanesh, S.Vijayprasath, “Analysis and Effective Implementation of Mobile Based Tele-Alert System for Enhancing Remote Health-Care Scenario”, healthmed Journal, ISSN : 1840-2291, Vol. : 6, Issue 7, pp. 2370–2377, 2012.
15. Shreyakamani,neelvasa,kritisrivastava,"Virtualtrialroomusingaugmentedreality",
internationaljournalofadvanced computertechnology(IJACT),Vol.3/6,pp.98-102,Dec.2014 16. M.Paranthaman, S.Palanivel Rajan, “Design of Implantable Antenna for Biomedical
Applications”, International Journal of Advanced Science and Technology, P-ISSN: 2005-4238, E-ISSN: 2207-6360, Vol. : 28, Issue : 17, pp. 85-90, 2019.
17. S.Palanivel Rajan, et.al., “Intelligent Wireless Mobile Patient Monitoring System”, IEEE Digital Library Xplore, ISBN No. 978-1-4244-7769-2, INSPEC Accession Number: 11745297, IEEE Catalog Number: CFP1044K-ART, pp. 540-543, 2010.
18. Kusumaningsihandekomulyantoyuniarno, “User Experience Measurement On Virtual dressingroomofmadurabatikclothes”,IEEE,2017
19. S.Palanivel Rajan, C.Vivek, “Analysis and Design of Microstrip Patch Antenna for Radar Communication”, Journal of Electrical Engineering & Technology, E-ISSN: 2093-7423, P-ISSN: 1975-0102, Vol. : 14, Issue : 2, DOI: 10.1007/s42835-018-00072-y, pp. 923–929, 2019.
20. Dr. Anthony L. Brooks and Dr.evapetersson Brooks “Towardsan Inclusive Virtual Dressing Roomfor Wheelchair- boundcustomers”978-1-4799-5158-1/14/$31.00©IEEE,2014masakiizutsu, and Shosiro HATAKEYAMA “Estimation Method of Clothes Size for Virtual fittingroom with Kinect Sensor” 978-1-4799- 0652-9/13 $31.00 ©IEEE, 2013. Ting Liu and lingzhi Li,“Real-time3dvirtualdressingbasedonusers”,IEEE2017
21. S.Palanivel Rajan, K.Sheik Davood, “Performance Evaluation on Automatic Follicles Detection in the Ovary”, International Journal of Applied Engineering Research, ISSN : 0973- 4562, Vol. 10, Issue 55, pp. 1-5, 2015.
22. J.Liu,“anewreadinginterfacedesignforseniorcitizens”,Instrumentation,Measurement,Computer,co mmunicationandcontrol,pp.349-352,October2011.
23. S.Palanivel Rajan, T.Dinesh, “Statistical Investigation of EEG Based Abnormal Fatigue Detection Using labview", International Journal of Applied Engineering Research, ISSN: 0973-4562, Vol. 10, Issue 43, pp.30426-30431, 2015.
24. S.Rigos,“Ahardwareaccelerationunitforfacedetection”,mediterraneanconferenceonembeddedcomp uting(MECO),pp.17-21,June 2012.
