SM DESIGN ICLAB Real es Video Survelliance Anomal ! z B:/ : lie | L Ee 5 f = zs 3 F ° ree CN eC} ie . GSmart Domestic Solar System DUCE CUM OUC UD COOMA a ‘School Of Electrical Engineering And Computer Science 1.PROBLEM STATEMENT ACMA MEM a atte mle opie eal frequency elements to achieve pure x70 AIM tle MUR CMCR te itd {86 YAN aal aM (erent ave ped efola Una for load management. a al a 2.BACKGROUND el af e| 4 Pentre kal Mobile App conte! fo. cells to Square or Modified sine waves ARUN] Naat ce Tear damages the appliances. Controllers for load managemt & user interface were Ore Ee tt ee 3.OBJECTIVE Modified Sine Wave To design an efficient system that pr Pee the load smartly 4.PROPOSED SOLUTION Pure sine can be achieved by using LC filter circuit. The Microcon: troller Unit will be programmed to manage the load smartly. We Bh MNCS rene erin releriauen okie Matas sao Me RCM EVN e- Mae NEMS E las e lM circ AMEN LU CoM colg ols IU ao) © PROTEUS ficite 6.CONCLUSION ARDUINO Cen Re ur omnes UNIT Ree ores ® Intelligent household load management during peak cost hours Renrenniercteoni ice PW UPelcleh alter e ceca EMEC umes Keou cohFarming has never been smarter! Design and Development of loT Network for Smart Aquaponics PROBLEM STATEMENT With population growth and water shortages on the rise, agricultural land is decreasing. How ca Integration of lOT into an aquap system for development ofasma More unique farming method resulting in provision of greater of crop and fish, i Recorded Time DATA IN REAL TIME METHODOLOGY I t eS Control and AQ System Block Diagram CONCLUSION Smart aquaponics system on large scale can help increase production of food while greatly reducing water and land consumption. Group Members: Advisor Syed Abdul WasayQamar (241685) Dy. Nosherwan Shoa Syed Jawwad Raza (244875) Co. Advisor: Rehman Afzal (218058) Dr. Hammad CheenPTC Acca Using NUST mm-wave Radar Oa eeu i a Cana) re Falling, the main cause of aN ea a Di disability in the elderly. » as tat One in four elderly people ‘oo ‘car assent falls, but less than half talk to ‘ | their doctor{1]. Requirements: > = 60 GHz mm-wave radar deve sect id Bee een or rc eesaion| |2. Conventional Newal Netvork ee + Raspberry pi oe EE > Data extraction with single ‘and with four sensor radar. > Development and comparison of different machine learning models. > Use of Raspberry Pi for Embedded solution and notification. > PC independent project. Figure |: Plot of taining and validation accuracy Figure 2: Plot of training and validation Loss Pt rd cl ¥ Convolutional Neural Network training with { fm a | 94 percent accuracy. ¥ Different radars module with different * Ke Radar ranges. B # ¥ Implementation with one and multiple sensor. CNN ef — Fall detection ¥ Raspberry Pi for an embedded integration. + Stand alone model + Fall event notification [I] AL. P. Siciliano, "fall detection and posture recognition in elderly healthcare," IEEE, 2010.Abstract Results Point position advances in technology has opened modalities of communication and human ter interaction. We expand upon this by Bounding Box Our objective is to use computer vision techniques to identify hand gestures and use a custom vocabulary to translate these gestures. into application specific commands Gesture _, recognition Methodology and Technology We have broken down our project into two primary constituents, namely gesture recognition and robot piloting. We have developed a model based on the Google MediaPipe pipeline for gesture recognition ean and will use a custom dictionary to first control the area movement of the Pioneer 3 robot. We will be taking our video stream from an OAK-D camera. After integration of both these, we will move on to the next phase which will be able to perform more complex tasks, such as a swarm operating in unity References é to move an object. H. Nishimura Et al. “Risk-Senstive Sequential Action Control with Multi-Modal Human. A of, cu itil e Trajectory Forecasting for Safe Crowd-Robot Interaction” OpenCV htto:/wwrw.giannidicaro.comMroboties.htmlThe 3D priniing Is one of the leading emerging tecinclogies of Industy 4.0 . The use and implementation of additive manufacturing, in ‘cor bination with other technologies, is producing an evolution in tha Industry towards an. intelligent procuction whera machines (autonomous, automatic anc! intelligent), systems and networks are able to exchange information and respond to the systems of production management. In addition, as a technology capable of tuming a 3D design into o product without intervention, 3D printing has o fundamental role, Moreover, the need for expensive tools and fixtures is eliminated, reducing post- processing, material waste and human intervention. These are characteristics that define the industry of the future METHODOLOGY In 3D printing, one creates a design of an object using software, and the 3D printer creates the object by adding layer upon layer of material until the shape, of the object is formed. The object can be made using a number of printing materials, including plastics, powders, filaments and paper. FDM 3D printing Is @ technology that works both horizontally cand vertically, where an extrusion nozzle moves over «build platform. The process involves the use of, thermoplastic material that reaches melting point and is then forced out, to create a 8D object layer by layer. As the design takes shape, itis clear to the see ‘ech layer as a horizontal cross section. Following ‘the completion of one layer, the nozzle of the printer Is highered in order for the next layer to be added to the design.APPLICATIONS 5D printing use is thot itso simple technology that can be used in applications in all kinds of fields. » Education '* Prototyping Ar * Construction * Manufacturing + Jewelry + Medicine © Construction * Research + Architecture * Firearms * Cultural Heritage. + Automotive © Spoce * Defence '* Domestic '* Aerospace PROJECT GOALS Fabrication of 5D Printer is startup based project. Our goal is to democratize the manufacturing industry, from food to medical supplies. We want to create future of 8D printing machines by making their way into homes, businesses, dleaster sites, hospitals, colleges etc. In order to achieve these gools we are Interested in making « space for makers and want to establish our own startup and then expand our business model into the following mentioned categories of 5D Printing. + 80 Printers Manufacturing * 5D Printer Farm ‘+ 80 Printer Filament Manufaturing + 50 Printed Items ‘+ 8D Printer Raw Materials * 3D Printer Filament Machines Manufaturing ‘+ 3D Printing Services + 30 Printed 3D Printer Parts» 80 Printer Filament Rew Materials Monufaturing MOM (ULE Ka Te aac: Ta nee | | and tre Abstract Our project uses deep learning methodologies to accomplish tw‘ further classification of for 9 main goals: Segmentation est areas of Landsat images into forest and non-forest areas and into different species of trees. Objectives Based on forest inventories, policymakers are able manage forests. Physically doing so Is jes like ours will help the governmen to make decisions to sustainably e intensive. Using resourc bilitate Pee ccc it in their efforts to rehal remote sensing technologi forests in Pakistan. Research Methodology Results aaa Ne Noa) ce yranee odd Fig:t Visualizing 2,3 Fig:2 Corresponding and 4 (r,g/b) bands segmentation map classification of Landsat nese of of forest and non the forest regions region Chitral forest area of Chitral . Region Related Literature 1 Chiang, SH; Valdez, M. Tree Species Classification by Integrating Satelite imagery and Topographic Variables Using Maximum Entropy Method in a Mongolian Forest. Forests 2019, 10,961 2. Bjerreskov, Kristian Skau, Thomas Nord-Larsen, and Rasmus Fensholt. "Forest Type and Tree Species Classification of Nemoral Forests With Sentinel and 2 Time Series Data.” (2021),Adversarial Attacks on Self-driving Cars Advisors: Dr. Faisal Shafait, Dr. Shahzad Younis Muhammad Umar Javed mjaved bscs1Bseecs@seecs edu.pk Muhammad Hasnain Naeem Abstract Deep neural networks (DNNS) ate being ‘used In vsion-based autonomous diving. tems, However, deep learning rods have Jpeen found vulnerable 10 adversarial tacks, ‘where small magnitude peturbations can cause ramatic changes 4 the outps, ‘This work ‘ims 40 Weniy and address physica wold challenges in generating adversata examples ‘nd (0 devise methods to generate robust a ‘eraralexaapls for physical word use eases ‘he aikonomoas driving system, Antroduetion etree ae a ae Titinigiafoeward tank, however mapping hve de ht advesaral examples tte veal wrk els ‘ows inelfectvenes because o fem challenges, Autonomous Driving ‘A deepen based autnomoas dasa stem is composed of te functional ayer, ad an ade inal lod serie lye from ea images by adding a small perturbation, To generate the optimal per sary may use a generative model tained Us Felowingoptintzation probe sgn 32 tad x) + Uap Sl Te objective I 40 mile the percep sin Atty low tetvee the adver mage and the ‘pital age an 10 maxi hs tose wet of there model Physical World Challenges |,One to Many Challenge: Ove advent ex ample must te sble mislead the ving ra mail viewing angles ogo the deving roves Figure 3: The same avers top sgn ss be able to mild the vehicle at ier a lated Areas The verry is ‘nly alowed manipulate smal secon of he ‘ete eld few othe storms vehi an only mani the igure 5: General Prost Worklow Results —PSGAN Swe | 197 Dave | ‘ines Wiss 95207-2557 ges | ase | “ae! | 3 ay] WIS | MET ety Rambo | 1160 | 751. | is. St ting angle MSE (et) and (ie for dierent secing ts une SAE wa vemos under ute Faure Goats “tenn he tans of (dite ging conto ose! ph “*Atlacking a wider class of Models. es tenn ater pero bak box tec esl example, References mae Suet PX l4126572. 2014 Arkiv pe tiptoe kal hm si tg, ig a onaton 5 nib and hanna Proceedien gp ference re attoncyRoam Sa Abstract web portal based on Mapbox that showcases the best travel and adventure destinations in Pakistan swith special focus on the offerings of Gilgit Baltistan. J community of adventurers and a rich database with Guidelines of experiences in Pakistan. Minimum Viable Product ‘Web based map-box react app that has match making gnvironment incorporating Sentiment Analysis using NLP techniques. Infrastructure Roam Server For ML Mode! Processing | iat, — django. —— & 2 Technologies — Bab B References [1] Ashish Vaswani, Noam Shazeer, Niki Jakob Uszkoreit, Llion Jones, Aidan N Ry Gomez, Lukasz Kaiser, and ilia Polosukhin. Attention is all you need. arxiv preprint arxiv:1706.03762. 2017. [2] Del Tredici, Marco, Marcheggiani, Di Im Walde,Sabine,and Ferrand, Raquel, shall know a user by the company it keeps: for social media ea a and Research Ml titision sAatad dinnnit Dynamic representations in nlp. 4701-4711. 10.18653/v1/d19-1477. 2ee me ABSTRACT ndern day ‘celivery. systenis cause carbon emissions, traffic issues, and require direct human input. Automating this system with an actonomous, green. flight system would be very beneficial to the future of delivery systems, as delivery-is becoming an increasingly common method of getting shopping done, be it for food or accessories, Many large companies are already looking into this, but a complete solution does not exist. Using tools like ArduPilot Mission Planner. OpenCV and Pixhawk, we plan, to make and program a drone that can deliver an item (like a document) from Point A to Point B autonomously without human input OBJECTIVES + Reducing carbon emissions + Reducing traffic in general * Fast delivery of urgent items (eg medical supplies) + Address and contribute towards achieving UN 2030 SDGs to achieve a better and more sustainable future for all at ‘ox mow 7 a a ‘THEGLOBAL GOALS: FINDINGS After assembling the drone, we calculated its theoretical payload capacity: Payload capacity = (A *B"D)-C Where A = Motor Thrust, B = Number of Motors C= Weight of Drone, D = Hover Thrust With D = 05, the drone can carry 400 grams With D = 1, the drone can carry 900 grams: AUTONOMOUS «> FOR DOCUMENTS AN Lic*t ITEMS DELIVERY PROBLEM STATEMENT Most modern day delivery systems cause carbon emissions and require direct human input, employee costs and are a burden on existing transport infrastructure, so automation would go a long way to streamlining this process. METHODOLOGY USER SENDS aa DOCUMENT DETAILS, > =a FACE 1D AND = DROP OFF LOCATION SERVER CALCULATES ROUTE USING MISSION PLANNER AND SENDS COORDINATES AND. FACE 10 TO DRONE CAMERA USES FACEIO 10 VERIFY USER, THEN THE DRONE LANOS AND DOCUMENT. 1S DELIVERED TO USER ‘ONE Is LATED WrTa DOCUMEPT AND FLIES TO SET LOCATION AND ASKS ‘USER TO POSITION THEMSELF INFRONT OF CAMERA, DOCUMENT RECEIVED ‘DRONE RETURNS BACK TO BASEwilling to sell the two. =i =s Search Advisors Chat Easy Payment Manage Bookings Reviews | rin MSc hs Need Advice? Find an Advisor BookaMeeting _ Pay and Review Sign up and Log in on) Search Advisors, Find | Negotiate timeslot || Pay for your meeting Helpex the right one for you | and book meeting leave a “ Development Tools Flutter nedew @ MongopB aws BT Terie) PET g Muhammad Yousaf Saddique Pla Paar ltt 4 Muhammad Talha Qadir Co-Advisor Muhammad Haider Ahmad Dr. Yasir FaheemVent Out Abstract A conversational chatbot application wit on a personal level, ensuring that no one c: Problem Statement * Expressing emotions can be difficult as there is a fear of being judged. Research has shown that suppressing emotions can cause various mental and physical implications. Methodology Audio converts User speaks to chatbot to Text Architecture Dr. Hasan Ali Khattak -@-@> es ML proc backer A BUDDY RIGHT IN YOUR POCKET Aims & Objectives + A user friendly that understands v jout any + On hearing the conversation that helps t their f Features + Voice-based conversation Text-based conversation Interactive chatbot avatar End-to-end encrypted conversation TechnologiesAKAUNTIN ey 03. PROJECT OBJECTIVES + Adding mobile interfoce + Adding double entry 04. TOOLS USED + Adding custom vis + Adding forecesting feature —— Ce ar 05. RESULTS + Business owner wil have oll moder features + Business idea by selling plugins for our odded feotures + Akounting will stand-out between competitors os for no other accounting softwore provide these fectures ot one placeP= -NUST oof tects Engineering ter Science (SEECS) Group Members; Talia Hamid-261832 Faizan Ahmed-245112 erernency services in Pakistan are provided by calting on separate heiyines. There is no such integrated cystem which svovides the ‘rovision of help from a single platform. To cater the need of a ed emergency service provision gystem a muoile apr is evelope* with Urdu Ul and Urdu speech recognition, which maiorily understands easily in Pakistan. We aim to’ make emergency services accessible to the cecple with ease and enhanced functionalities at one platform, 02 tee hha + In emergency, people are usually unable to cec/phat what to do and are not able to type + Most of the times they don't know the exact location + They forget the helpline numbers of the relevant department ll to a helpline is not free 03 This project aims to provide an android application through which, ‘emergency services can be provided to the users. It includes: ‘+ Emergency report input from the user through text and speech inurdu ing the location ofthe user on of necessary help guidelines til the arrival of help, ony he ole reared using NU scion Fetching Usha BS Emergency Services App Advise Dr. Ratla Mumtaz Co-Advisors: Or, AU Tahir Sir Maajid Maqboo! 05 CE OS ‘The result ofthe survey conducted to decide top three functionalities of the app is shown below. “O° ered Jo Horne Bereen UN O7 ‘The app thal we are laaking 1 devele vilations What Ihe pre-existing siiviar satway (woking forward (6 the leaning louiey ant challenges that come our way doing the FYE have We aeDeep Learning for Semantic Segmentation of Satellite Imagery 1) Abstract 5) Architecture Satellite Image 2) Problem + “The glaciers of the Himalaya would be gone by 2035 (IPCC 2007, chapter 10.6).” -World Bank. >< Neural Network Glacier’s melting would 7 Model Lead towards water scarcity , a in Pakistan. of glaciersis difficult, so this method effort of collecting glacial information, tof satelite data can give moreU-ASSISTANT ABSTRACT We plan to make a virtual assistant for smartphones that understands Urdu. The target audience will be Pakistani coins users who may not be well-versed witi technology and the English 1anguaxe. CEN Rea ay, : U-ASSiSTANT Group members: Muhammad Hamza Amir, Abdul Rafay Ahmad Advisors: Mr. Taufiq ur Rehman, Dr. Muhammad Ali Tahir ROBLEM STATEMENT « The majority of people in Pakistan living in rural areas do not have access to the rroner education systems, thus are not weil }crsed in English. + The Usage of smartphones is increasing ar, but some people find sinartphones difficult to use without every proper guidance, *Al assistants are widely available on smartphones and the users can perform, many tasks efficiently by simply issuing a voice command. *This project is aimed to deliver an Al Urdu voice assistant for smartphones, which shall be readily available for Android users. PROPOSED SOLUTIO Building an Urdu speech-based Al Assistant for smartphones that should be able to: * Set an alarm + Make phone call + Add/ remove contacts * Open applications * Show weather METHODOLOGY Speech to Text Text Generator AO Input Denoiscr Output a Text to Speech ao Text Analyzer NLP Mobile API ey PROTOTYPE ®O fF [RASA] Nw¢ PROBLEM Bridge aims to solve the problem that, a large number of the social media users, businesses and content creators are not aware of the availability of the solutions to handle the social media platforms easily. Even if they are aware of it, they are not using them due to very limited features available in free versions and high prices of the premium versions. Some find, them more complicated to use due to less documentations and trainings available, KEY FEATURES ‘Management of social media platforms through 1 app. Providing services like scheduling, hashtags and notifications 1. 2s 3. Helping businesses to grow and save time 4. A cost effective and user-friendly social media ‘management platform ABSTRACT Bridge, a cost-effective web application that would assist a novice user to easily share and schedule posts for multiple social media platforms like Instagram, Facebook, Twitterete. It is primarily aimed to assist business enterprises of any size/ magnitude and content creators to handle their content(s) effectively and conveniently on multiple social media platforms from a single platform OUR CONTRIBUTION *v° EXISTING SOLUTIONS: Hootsuite Latermedia (OUR CONTRIBUTION: Providing User Manual and Guidelines ‘Scheduling Notifications Preview Options Hashtags and Highlights > DEVELOPMENT METHODOLOGY Using Agile ( SCRUM }, for the development. Repeating steps of requirements, design, code (frontend and backend, integrating and testing for each module. The block diagram for the features of the project is given below: Treas (Connect Socal fecisertcaieny) cae as) Handle a ay SU esncatna)) ak Matstone) Scheie Pox) Bae “hdd Library & _/ Add Payment > Highlights Feature Abasor Sit Taleeg:x-Rehman Co-advisor Si Masjid Maqhool RESULT ‘Asocial management platform to post anything in an instant ‘of time to all other social media platforms to save time and better user-friendly experience. TOOLS AND TECHNOLOGIES 3& Expressis React ‘Mongod8 Tom Members: Anum Mujahid Fatima Khai, Ghazals Bibiaor Abstract Introducing come automated assistance in cooking might help to solve the cooking-related problems where the chef would have to just issue the commands and the assistont will provide instructions through speech ond text support Problem Statement The real struggle while cooking starts when you follow a recipe and in between, you realize thot the ingredient is either missing in the dish or in your kitchen. Following @ recipe tutorial is aleo one big hassle; if you miss © step oF ingredient you have to look for it manually in a recipe description which might not be possible with dirty hands full of cooking ingredients 22> & Our Solution Pe etch tom nen Reet Scien cs one ee ee aa cen, ee ic cmcm all the way to make a delicious meal. @ RecipeBot Virtual Chatbot Based Ariston! for Cooking Methodology ABA t — oe ‘ Modeling Deep Neural Networks Data collection Data preprocessing User Request a =p Assatance Throughout Cooking Response Generator Results ‘Survey Findings Forms o Recipes Fllowed \ = \ ~ a } cal : Sources of internet fllowed for recipes Will irual Assistant for dictating recipes make life easier? Conclusion Cur system is an intelligent system with the capability of communicating to the user through speech or text. It uses @ deep. noural network to provide a user-intended response, Thus, i will bo a useful application to the users to give them « hassle-free cooking experience. @S & panthon Deep Neurat Network React mongoDBADJECT OBIT MENT METHODOLOGY PROJECT OBJECTIVES DEEL EMENT Face recognition, feature extraction, and object detection modules have various (Detection Systems] fection Systems algorithms with high-performance pen SOP VIOLATION Mahe ape cs e DETECTION ‘extensive public application. 1 : PleiittrM p Sr || m| Ig EAS THROUGH LIVE IDEDS User interface “Two major modules; Face Mask and Person Detection are pillars of the intelligent system APPLICATIONS i -Elee S41 Er, facturing and 3 Creracal eseticn &, os Healthcare: OPDs, 5% ha Boon X ICUs.Nurseries Seca ea |» Medicinal Research Laboratories CULT) 4 Public Places: Bank, Supermarkets, f : extieg Educational institutes. : REC EES Rm AMAL) s MTC eel crust ey ‘otection Distance eae cues jicinities are the 7 5 itical requirements. . corer si WAL Re eS ou ae ku cre Bernese) COS Cee Ru mu | ironments PMOL GLAAL Ud Cem NTS Peat) Cine make anaes7 SLUW Ve ON cele Re tell Molar ae Leu (cd forte Mero ch Lola) Background Citizen complaints go unresolved due an's ability as a result ent d sharing. spreading es an integrated framework consisting of 3 web-based dashboard forthe state run departments, ‘smartphone applications for citizens and government employees. Sahulat, with its social aspect and smart complain management features, provides the perfect solution to bridge the massive gap between governing bodies and its citizens. ce SI ae WUT PLL ra Ey Y EC) ahulat’s Methodology Individual Municipal/Civie Complaints. Issues 8 m— EVERYBODY BENEFITS | Project Goals % A o = . So Tools & Technologies ® BG myso ® camedElectronic projects involving motors and actuators can have a vastly different hardware and control requirements. Currently in market, there is a lack of “single board solution” for easily working with different motor types. This project aims to produce a configurable board that provides multiple simple interfaces for controlling multiple brushed and brushless DC motors. Our development methodology is based on red: of cost and size by minimizing hardware and writ airtight firmware to make maximum us microcontroller's resources. + The objective of this project is to produce a board with the following feature tree on each of the four channels: + The channels will be controllable via UART, SPI, 12C, servo PWM or an analog signal. = Quadrature encoder, simple rotatory encoder and analog feedback would be possible. + Selection between available features will be managed via.a GUI program. ‘Automotor is intended to be at the heart of the motion in applications such as: + Multicoptors and Drones * Robots * Electric Vehicles ‘+ MPLABxIDE and STM32CubelDE are being used for programming. + AUtoCAD EAGLE is being used for schematic and PCB designing. Our final outcome is a multi-modal, multi-channel DC motor controller which is configured using a GUI program. In conclusion, our product: an drive both brushed and brushless DC motors. Has integrated PID loops for speed and angle/position control. Can be controlled using any of a large number of protocols. ‘Allows selecting between the large number of ‘options through an easy and intuitive GUI. Can be used to save cost, space and time in a myriad of application areas.ON THE EDGE EMBEDDED FAULT DETECTION IN ELECTROMECHANICAL SYSTEM Inds Machine Life equipment requir and cycle of maintenance. The — internal nd faults in the system ean atastrophe and big loss, OBJECTIVES We have to develop an embedded system that uses onboard machine learning to predict and detect if the Electro-mechanical System has become faulty or not. Group Members: — Uzman Munib Muhammad Abdutlah “Syed Muhammad Kamran Shatgat METHODOLOGY Wiest, we with employ: Machine hoarning It Deep Neural Network for feature mamuing in fault dlagnoala of electramochanioal aystoms, O Nae propane Pirexach is 40 Use variability of input Sensory «lata te proved fant DNN based! alyoritiin oth are analyaenl seWRE public onvain dataset for nat AN Lovet Mawet for evaluation of anomaly d tection algorithms eS a weldled systoms are usedl for time validation, Nvidia Nano Jetwon SYMA2E29Z1 Comtex Ma ® node) With vibration, citrront nel temperature sensor a2 i i- iv “a ‘quisition system on Ataluine Uno. RESULTS We have tested our clawitication and jreadietion embedded MCUs, We have aloo evaluated different atonclon alyaritlinnn the network takes different hardivare plattorays Holow are vane renults that we achieved nt catnsot, [won| ENO wh CONCLUSION Compared — tw tractiti Processing, paradigas, inherently datardriven, provides and energy efficient plat leading, to a robust faull diagnosis method Advinor Dr Usman Zabit Co-Adivsor Dr Wajid Mumtazexpensive to be performed devices. O: feismices { taxing part Paice oso) recognito es Ceol ith or will aes peed +, a nae = APPLICATIONS ii) @tualiiyreseurance at call centers 2) Automatic call monitoring 3) Used as an input stage for sentiment analysis AlsaT @) Real-Time Unmasking of Masked Faces RISC Cees Ore rs ei caheddhteecinterh ch ieee Latin lh aehecenainil 4.BLOCK DIAGRAM Perri od Pr Cay Oe ad Building a Masked Face Recognition Deep Learning Cen) [ 2.AIM i rT i Rit RPE Pn ocala aad images because itis a very intriguing problem of great practical value. Model Training Performance Analysis ACKNOWLEDGEMENTS ea iat a A eat TheFa ADVISORS DANCE IEC i a cs Meee PR Sam ue Lee Bede cesar e leer) laa Ue sl imu eels a ue ae is Keras F tensorriow B-- Pon Dr. Rehan Ahmedwt vou Crop Disease Scouting RA Es on Using Al & lol Technology iy Abstract eu W Regge adenine: ae f * To. reduc c ground surveys for crop © To develop a web por ze, manage the data and perform ad} ution using Cloud and Edg Methodology unted on Nvidia Jetsor of 96.196% was obtained t g U2-Net Data Data Model Image ResNet-18, The resul Collection Preprocessing Training urce dataset Classification Segmentation on Wheat Stripe Rust Disease Classification using Computer Vision and Deep Learning paper published in IEEE A Pao Ce eae Cy ErreurEx Information Extraction From Email Abstract Development Methodology MailEx is an Information Extraction tool + MailEx provides a Server which: WebApp which the Information * Uses Artificial Neural Networks (ANNs) users can use to | 9" | to extract information from emails upload emails and = Provides the extracted information to download summarized user in a database such as csv file information NLP techniques employed 7 oa = 1, Named Entity Recognition (NER) ie 2. Relationship Extraction (RE) + Tools used 1. Tensorflow and Keras for NLP Background and Objectives 2. React and Django for WebApp Contemporary Solutions 3. Azure Cloud for Deployment + Follow Rule-Based approach + Are narrow sighted + Cannot detect context MailEx + Uses Artificial Neural Networks (ANNs) + Capable of detecting context + Extracts information accordingly Features + Allow users to provide mails forwarding or through WebApp + Provide output as downloadable web attachments + Secure data transmission + Ensure privacy by ignoring sensitive information Future Work Evolution of model to proces text documents suc3) 1122-Al A Chatbot based Emergency Hotline ABSTRACT Response time during critical emergencies could be longer than expected in real time situations. 1122-Al is an Al based ‘Chatbot Application that interacts with the users in natural language, extracts and sends useful information to emergency Tesponse dispatchers, tracks and coordinates timely help arrival, reducing the impact of the crisis. PROBLEMS tel certs eee na coy ~~ Arapid increase in number of public emergencies Musas eins E Shear volume of calls straining rescue call centres cs neem te ees © _Time-sensitive calls being put on hold STM piles eden ars ct ee METHODOLOGY AND APPROACH Logistic Regression Mode! Must Label Emergency Emergency Services Cassfcaton—_, Emergency ee : ms | Sener SULTS SURVEY MULTI-LABEL CLASSIFICATION OF EMERGENCIES USING LOGISTIC REGRESSION Preferred Mode of Important Features Communication ere a ee ie reaiee eae Ren scene nl Be mitre ice ea14% >] 61.28% omoe peters eaet oo = SS = css DISCUSSION AND CONCLUSION The Al primarily extracts and prioritizes calls for dispatchers while the crucial decisions stil rest with the dispatcher. Hence, it can be integrated into existing emergency response processes without distracting from or disrupting workflows, Advisor Bee err eis TE aescash (C2) er eeu eters ents and with digital payme ml (C2) js the first truly digital wallet which enables instant transfers makes investments accessible through eryptocurrency. | ero [ Tools and Technologies J Q] ee Conclusions and Discussion ; Ga) GE) A case study will be conducted with CryptoCash has comp! cece ten oes ry ] accept QR payments through 0) preety Deson | 8 vil ort ana ae * C2 functions as a pay to solve the lack of service for bu: ment gateway ' 8 good payment nesses in P: ‘ ee ateRSE In the field of homeopathy, there is Jumbled-up data of many remedies against each rubric without any experimental or theoretical proof in different repertories which causes problems for inexperienced homeopaths. UAT eT * Create an easy-to-use platform that collects the data of remedies, along with their effect, in real-time from experienced doctors, which were proposed to the patients. * Analyze the data collected and suggest the most potent remedy against a rubric. OCT ‘Data exertion from repertory Decabase Creation (=) BE completed TB inprogress I Pending ns & @ ® o. © Homeopathy is still preferred extensively in Pakistan, but people don't trust them due to the inexpertness of its doctors. Our System will help those inexperienced homeopaths in building their trust amongst their patients. ¢ Furthermore adding speech recognition and Urdu translation of the web app will help users with weak respective rubric. (s English in finding the most potent remedies against the Sir Taufiq Ur Rehman Kiani Dr. Hashir Moheed CATS 1.Muhammad Ammar 2. Muhammad Waleed 3.Muhammad AliDia tae yan * In the recent years, there has been fon increase in the amount of information available, especially on the internet. * Due to a lack of time, there is o need 10 summorize oll this information. CG) “Methodology. We have divided our methodology into 4 modules (Fig. 1): * Text Summarization + Interface © Query Interpretation a cia cree (@ ‘ We have conducted experiments ‘on the CNN/Doily Mail dotaset, hich is the most popular dotose! 33 for summarization. The Tools HSC MSG. Thslocing ROUGE ROUGE2 ROUGEL We aim to develop an application that con enable a person to search for ony topic and get the required information in a summarized manner. The goal of our project is to create o ‘system that could give users the information relevant to their topic in ‘a summarized form. ea uit otProblem Statement + Android adonts an open-source policy, + It also support unofficial third-party app stcres. + Its easy to release cracked appora normal application repackaged by a malware. How it woiks e@- User Downloads an app to install decompiled Features are Extracted from Manifest file Conclusion + Using the Machine Learning model we have designed an app that can effeciently detect malicious apps on user's device. eID Ti sty oH Saif Ur Rehman Ee) AndroSecure Android Based Malicious Application detection using Machine Learning ML model scans for malware Features are sent to. ML model for scanning Solution + AndroSecure, a Malware Detection App. * It scans the android package files before installation + Detects malware by analyzing the permissions used by the app. Scanning Complete Ifappis safe installation starts If threat found user receives a | warning message Tools Used Del Dr Mehdi Hussain Co-Advisor: PIR OL TayaAnnotation Tool for Biomedical m Statement Medical institutions generate large volumes of biomedical image data that can be used to tr models for valuable Al-assisted diagnosis in healthcare. However, annotating biomedical imag, process that also needs input from professionals for accuracy. Proposed Solution Objectives © Online, collaborative tool © Specialised for medical professionals &® PP Smooth completion of annotations with an Bf optimised user interface nyse ne © Human-In-The-Loop active learning architecture romana built on a pre-trained machine learning model e Methodology mapeee Human-in-The-Loop Active ee Learning System eo .. Hh. — > sea Be oer annetation Fle Data Flow Features @) seam" fa cevsvete Sa cent Human-In-The-Loop Machine Learning Technology ert yur: | Crear ea Caner ene na Ye eRe Le supervised Lima Conclusion The project will be evaluated on. the following metrics: @ Accuracy of the annotation prediction confidences © Time-taken to annotate compared to similar software © Evaluation by Nielsen's 10 Usability Heuristics\UDITORSPAL A PROJECT MANAGEMENT SYSTEM FOR AUDITORS PROBLEM STATEMENT + Auditing is one of the most essential requirement of SECP in Pakistan for registered companies. * Getting the finances audited through auditing firms is very costly. Small businesses cannot afford to get their finances audited. + Small businesses need an alternative way that would cost them less to get their finances audited. PROPOSED SOLUTION + Auditorspal aims to solve this problem by creating an online _ project management system for auditors. + A way to connect the auditors and the organization management directly without any auditing firm in between. + The company will not have to pay huge commission to auditing firm anymore. + The progress of the audit process will be shared with the company management on a regular basis. WORKING we ES Organization Managers Financial __statements will share basic and other required information about the documents with the business and they will auditor by the company get the services of management who ‘auditors with expertise wants to get their in that kind of business financial audit done. ‘Auditors will be able to Auditors will share the see all the audit final audit report with projects he/she is the ‘company's. working on and the manager. and he will company managers can give his reviews see the progress of the regarding the services audit process Provided by auditor. FEATURES Fig v Efficient Data collection Easy Communication Check on Pricing Easy to track Progress process TOOLS TEAM MEMBERS ADVISOR CO-ADVISORS B Figma Rao Obaid Ur Rehman Hasan Ali Khattak — Panlid Maabool Muhammad Waleed Asif Yasie akan 7 Kh Asad Mukhtar FlutterReh preva ee Pm SMa NUST Defining Futures AL-TABEEB BECUASE, WE CARI Abstract Healthcare records are increasingly becoming digitized. As patients move around the healthare ecosystem, their electronic health records must be available, discoverable, and understandable. Furthermore, the security of the system is our main concern, we will be using FHIR: enabled Healthcare Information Management System. Objectives Avoiding Document based Data: Feed Information directly into the provider workflow, avoiding pitfalls of Sooument-based exchange. Security of Data: We will be using —_Blookchain )) to secure our dat (Hyperted Sir Methodology laboratories, and hospital ‘their medical records easily. Vouk Nalevate Problem Statement Data interoperability is based on dooume! It can be too limiting for meaningful ca ‘coordination, decision making, or data analytics. nts. Document-based exchange of Dat: You haven't told the story of the patient, and fly important for a clinician to Patient data Is hard to collate: ‘When patients hi nt cat it's tough for physicians to collect and review past dat providers, Tools Used ° React nedes vevpentencer I) HL7 FHIR Conclusion Our project will base on a react website which will connect to blockchain and cloud using Nodejs and FHIRUI will be based on HCI principles. Group Members + Muhammad Hussain MurtazaWA Dark Web Forensics SOLUTION PROBLEM STATEMENT DEW AL CUE aang aL Sen eae ay Cae 9 + Identity Thefts Or Gir trys Pree tee eat TAN HELGE TAS Ce A aa (Peer mC ten acm Per Neus cece os Pee das Pia) Caeares eco Analy Cao eo) AMM 1 ese + — ao Q Wee i Daa eos ete 4 etecti PC a} eee TTT a Dea este) A ets ee 5% Displaying c 01 isplaying data CONCLUSION on dashboard senor i (Available to all) 3 Platform to see trends of DICE cree aE Te cao rte ares 9 } + G Lele ter nyc) RS St TET Can be used by law agencies websites d ec bat) Foyle - ets aa UE 4 Wate Testa ty DS TET) 3 Da eScCUM Ellis