Thesis (Bachelor of Science in Computer Science)

Browse

Now showing 1 - 20 of 885

Supershop management Models: An optimised way to manage the supershop using hyperautomation and machine learning
(BRAC University, 1/23/2023) Ahmed, Shuvro; Mojumder, Rajesh; Rahman, MD.Mahmudur; Karmoker, Joy; Fatin, Shadman; Rabiul Alam, Md. Golam; Reza, Md Tanzim
Customers are the heart and soul of supermarkets, and these stores often suffer losses due to mishandling of customer service. This study aims to examine how the satisfaction of customers can be maximized with the help of hyper-automation technologies in order to operate the stores successfully. Here, an intelligent voice bot is used to reduce response time for basic customer queries by providing real-time replies using NLP and for further complex queries, customers will be provided with contact info or forwarded to relevant authorities. In addition to that we predicted customer demands for future products by using multiple machine learning libraries like XGBoost, Linear Regression and Random forest with the help of daily sales data. This helps supermarkets to stock a perfect amount of products in their in ventory which will help them to avoid any kind of product shortage in any season and will help them to achieve customer satisfaction. Furthermore, optimized prod uct placement will be ensured with the use of data mining techniques like Apriori algorithm ,FP Growth algorithm and GSP algorithm. By doing this we are making it easy for the customers to find out their preferred product together in a single shelf and reducing customer hassle of iterating through the whole shop to find the products from their shopping list. As well as, to ensure a hassle-free transaction between consumer and seller, a system is proposed using Smart Contract System via Block chain which can make the process faster, and ensure transaction safety at the same time.For results, With an R-Squared score of 0.963, we discovered that the hybridization of linear-boost regression was the most suited for forecasting. The best outcomes for product placement were provided by FP Growth. For the pur poses of the chatbot, let’s say that for the two strings ”rfl nipple 3-6 month” and ”rfl nipple 3 to 6 month,” the spaCy llibrary and nltk’s bleu function both yield 90.8 and 66.21 percent similarity, respectively. Now, based on the %, you could assume that the spaCy library is operating more effectively, but this is untrue. SpaCy library functions. better in a big model where pre-trained word vectors are present, but not in a small model. However, the smart contract system successfully carried out all of the system’s algorithms and guaranteed transaction security as well as product safety. By using the Hyper Automation technology Super stores can ensure better customer service. The budget and implementation of such technologies combined into a system turned out to be both high and complicated. However, as time passes, the cost of these technologies will decrease rapidly and their usage will be further simplified.
An analysis on the effects of parenting style on offspring’s behavior using machine learning
(BRAC University, 12/4/2022) Akter, Nasrin; Mostakim, Moin; Reza, MD Tanzim
Parents are usually the most important person for a human being as they encourage and support an offspring’s physical, emotional, social, and intellectual development from infancy to maturity. An individual faces various challenges as they grow up. Proper parenting plays a prominent role in handling and abating those challenges. This paper aims to show various consequences on the attachment style and handling of depression, anxiety, stress, anger due to different types of parenting style. These consequences of parenting styles are to be figured out in an automated way so that one can acknowledge these factors on their own and bring various positive changes to their parenting. The term ”parenting style” refers to a collection of tactics that have various effects on children. These methods can have an impact on children’ minds that lasts long into adulthood, both positively and negatively. This research makes use of machine learning algorithms in order to differentiate between various parenting styles through various aspects of their life such as stress, anxiety, depression, attachment style, anger management etc. The lack of publicly accessible data prompted us to compile my own data set, which consisted of 2206 survey responses from students(school, college, university). Afterward, the survey data was stored and pre-processed. Then, machine learning algorithms such as Decision Tree, XG-BOOST, KNN, Support Vector Machine and Random Forest are utilized to detect parenting style by analyzing the effects of parenting on their offspring and the accuracy of these models are 84.70%, 76.71%, 87.30%, 87.30% and 85.185% sequentially.
Internship in IT audit
(BRAC University, 12/4/2022) Lateef, Rushnan Faraz; Morsalin, Talat; Alam, Md. Golam Rabiul
Human capacity in the auditing field is immense. The task not only ensures company security but also the safety of its employees and clients making it one of the most crucial sectors. To identify the lack of security and policy maintenance failures, auditing is necessary. Regular audits ensure the maintenance of software, hardware, office conditions and safety protocols which are essential to running the company smoothly. During my 6-month internship at NCC Bank, I was tasked with observing and conducting audits in various branch locations. The experience was enlightening and truly a valuable learning experience. Not only was I placed in an environment dealing with real world tasks and concerns, but the work environment itself was supportive in preparing me for what to expect in this field. With the opportunity to shadow meetings and my colleagues, and accompany them to various auditions inside and outside Dhaka, I witnessed lacking in the bank’s oath to uphold company policies which ensure data security and employee safety. It was crucial to identify these issues and report them accordingly step by step so that they could be rectified. Additionally, I had the opportunity to provide my insights and recommendations which were appreciated with ample constructive feedback. Undeniably, there are some aspects in the auditing sector which cannot be digitized. For instance, visiting locations, seeing the condition of the offices, talking to employees and making detailed notes. While a hands-on auditing experience was crucial to my understanding of the real world and what this sector holds, an effective storage method is needed for reports. This inspired my app Report Manager which would house reporting activities and keep track to information being sent and received eliminating the need for any middlemen or manual labour. With further development, this app is targeted to better the workflow and reduce the stress that the teams might face during their reporting process.
Comparative study of X-ray and CT scan images for the detection of COVID-19 using deep learning
(BRAC University, 2015-08) Niloy, Ahashan Habib; Shiba, Shammi Akhter; Fahim, S.M. Farah Al; Faria, Faizun Nahar; Rahman, Md. Jamilur; Parvez, Mohammad Zavid
Coronavirus 2019 (in short, COVID-19), originated in the Wuhan province of China in December 2019, has been declared a global pandemic by WHO in March 2020. Since its inception, it’s rapid spread among nations had initially collapsed the world economy and the increasing death-pool created a strong fear among people as the virus spread through human contact. Initially doctors struggled to diagnose the increasing number of patients as there was less availability of testing kits and failed to treat people efficiently which ultimately led to the collapse of the health sector of several countries. To help doctors primarily diagnose the virus, researchers around the world have come up with some radiology imaging techniques using the Convo lutional Neural Network (CNN). While some of them worked on x-ray images and some others on CT scan images, none worked on both the image types. Thus there’s no way to know which image works better for a particular model. This, therefore, insisted us to perform a comparison between x-ray and CT scan images. Thus we came up with a novel CNN model named CoroPy which works for both the image types and shows that in 2 classes (normal and covid), CT scan images show a better accuracy and it is 99.17% whereas it is 95.73% for x-ray images. However, in the case of 3 classes (normal, covid and viral pneumonia), x-ray images show a better accuracy and it is 92.45% whereas it is 68.81% for CT scan images.
A new multi robot search algorithm using probabilistic finite state machine and Lennard Jones potential function
(BRAC University, 2017) Khan, Md. Shadnan Azwad; Hasan, Mohammad S.; Ahmed, Tarem
Swarm robotics is a decentralized approach to robotic systems. This paper exammes the problem of search and rescue using swarm robots. We present as solution a multi-robot search algorithm using probabilistic finite state machine and interaction inspired by Lennard-Jones potential function. The approach utilizes a finite state machine to separate the tasks performed and to change coordination rules according to the circumstances and social probabilities. The approach is tested in various scenarios to test flexibility, scalability and robustness. The performance results are promising and comparison with Robotic Darwinian Particle Swarm Optimization and Glowworm Swam Optimization for algorithmic complexity appear favourable.
IoT security risk analysis
(BRAC University, 2017) Sakib, Nazmus; Jerin, Ismot; Khan, Nuzhat; Quader, Shaela; Chakrabarty, Amitabha
Internet of things (IoT) has become a buzzword in today’s world to describe billions of devices, interconnected via the web. It includes a diverse range of devices, starting from wearable ultra low-powered gadgets like fitness bands to medical instruments and to home appliances to automobiles. There may be so many devices harnessing the power of IoT, however, security is still an issue for these devices as these are constrained with limited power supply, processing cycles and memory usage. The IoT sector is not impeccable, security is still a threat for the IoT devices as these devices are meant for low power usage for small scale setups. Security algorithms are not abrupt, but many of them don’t fit the IoT systems as their compatibility can only rely on products with larger form factor (Which usually means better performance, storage etc.). In such context, working with security of IoT devices has become an interesting area in computer science. Though, researchers and security professionals have developed advanced algorithms for ensuring digital security, but many of them are not suitable for the IoT world because of the restrictions we have. In our work, we tried to contribute to the matter of security in IoT devices. In this work, the main concern has been to investigate the performance of different security algorithms and compare them in terms of processing cycle and execution time in Raspberry Pi. We have worked with FLECC_IN_C and Crypto++, two different libraries with number of algorithms where we can find ecdh, ecdsa, ciphers, message authentication codes, one-way hash functions, public-key cryptosystems, key agreement schemes, and deflate compression. and measured their performance in a constrained environment. It is the first of its kind, to this work’s knowledge, to use raspberry pi which is established as black box device and implemented security algorithms on it. We implemented these libraries in different IoT platforms, showing comparisons of how these algorithms may affect a system in terms of resource utilization. The work in the end shows a summarised view of several key algorithms and decides which is better in the terms of IoT constraints.
Integration of data clustering to intelligent email classifier
(BRAC University, 2017) Zisan, Shahriar Ahmed; Moin Mostakim
In recent years there are more than billion of email users .People are dependent on electronic mails. ft has become an urgent and crucial component of communication .One of the cheapest and fastest communication mean. Emails can grow at a large scale. Sorting the emails according to their content can be a useful and time saving. Cluster analysis can come in action while classifying the email documents .Cluster analysis is a sub-field in machine learning a sub field of artificial intelligence, that refers to a group of algorithms that try to find a natural grouping of objects based on some objective metric. In general this problem is hard because a good grouping might be subjective. The k-means algorithm is one of the simplest and predominantly used algorithms for unsupervised learning procedure clustering. Existing email clusters are supervised clusters. In this thesis, we inherit the simplest machine learning strategies like k-means, k nearest neighbors and represent an intelligent email classifier combined with both supervised and unsupervised algorithm that categorizes email documents based on the body contents.
Comparative analysis between machine learning algorithms in efficiency of Coronary Heart Disease (CHD) prediction
(BRAC University, 2018-12) Oishi, Fayza Rezwana; Al Mahadi, Mehnaj; Parvez, Omar Bin; Arif, Hossain
The world of Machine Learning is expanding everyday through its implementations in modern day healthcare. Researchers have sketched out many ways to implement Machine Learning algorithms and droned into ways to make them work in their utmost efficiencies. As there will always be the need for healthcare in the world, we believe that there will always be a need of comparison between Machine Learning algorithms in terms of their performance and relevance to make healthcare more reliable through Machine Learning. For this study, we have picked up the most commonly used Machine Learning algorithms, Logistic Regression, Support Vector Machine, Decision Tree and Random Forest to produce a comparative analysis on a dataset of Framingham Heart Study which is dedicated to the prediction of risk of Coronary Heart Disease (CHD). We have used a combination of Data Preprocessing and Feature Selection methods, namely The Row Elimination method and Recursive Feature Elimination respectively. To understand the impact of each prevailing features in the dataset on the target feature, we have applied the Chi Squared Technique which is a highly recommended technique when it comes to classification problems. To compare and analyze performance of the algorithms, we applied concepts of the Confusion Matrix, Precision, Recall and F1 Scores; we have plotted ROC curves using Sensitivity and Specificity scores to categorize the algorithms’ behavior. We have found out that the highest average accuracy in our study was given by the Logistic Regression algorithm (83.9%) while the other algorithms have come fairly close.
Finding habitable exo planets using boosting algorithm
(BRAC University, 2018-12) Rahman, Md. Mashfiq; Afrin, Naba; Majumdar, Mahbub Alam
The first moment man extended their boundaries outside of our Earth, from that moment they were looking for another habitable planet, where they may live in future. Including NASA, many international space organization already sent a number of satellite on this mission. These mission have discovered thousands of new planetary candidates, many of which have been confirmed through follow up observations. A primary goal of the mission is to determine the occurrence rate of terrestrial-size planets within the Habitable Zone (HZ) of their host stars. Though many approaches have been taken to confirm their habitability, we tried a new approach by using boosting algorithms. We use the NASAs Extra Solar planets dataset of 3,577 planets and use their various characteristics like their Mass, Radius, Orbital Eccentricity, Temperature, Metallicity to determine the best set of alternatives of Earth. We classified the dataset based on these variables and used Extreme Gradient Boosting to compare the accuracy to find out our desirable results. We used different classifier to ensure the best accuracy. So we used Ada Boosting Classifier, KNeighbor’s Nearest Classifier (KNN), Gradient Boosting Classifier, Decision Tree Classifier and Random Forest Classifier into our dataset.
Modelling option prices using neural networks
(BRAC University, 2019) Nasim, Ahmed Zohair; Syed, Shehran; Majumdar, Mahbub Alam
In this research, modelling of the European option prices of S&P 500 index options was carried out using Multi-layer Perceptron Neural Networks. The goal was to train the neural networks using historical data to accurately determine option prices, given the index price, strike price and time to expiry as inputs. There is no hard and fast formula for pricing options, with the exception of the Black Scholes model, which is only a theoretical model and often under-performs in practical applications. Therefore, developing a model for pricing real options is of great importance, and Neural Networks have the potential to be vital vehicles to that end. That is what motivated this study. Di erent results with respect to accuracy are achieved by partitioning the data according to moneyness of options, with the Neural Network performing exceptionally for in-the-money options, but poorly for out-of-the-money options. This suggest that in a volatile market the neural network outperforms the Black Scholes model for in-the-money options, however the Black Scholes model is still better for at-the-money options.
Detection of amyotrophic lateral sclerosis using signal processing and machine learning
(BRAC University, 2019-04) Ali, Zohair Mehtab; Parvez, Mohammad Zavid
Electromyography(EMG) signals provide signi cant information for the diagnosis of neuromuscular disorders like Amyotrophic Lateral Sclerosis(ALS) which is a form of Motor Neuron Disease(MND). Due to the stochastic nature of EMG signals di erent preprocessing and feature extraction techniques need to be applied in order to extract useful information from the raw noisy signals. Time-Frequency analysis and EMG Decomposition are two of the widely implemented techniques for feature extraction from EMG signals. However, due to extrinsic and intrinsic artifacts any one feature extraction technique alone does not provide enough information in order to show a consistent performance of classi cation across a variety of dataset. EMG signal data set acquired from di erent sources provide varying outcome when passed through the same classi cation technique. This is a major problem while creating software which is able to perform automated classi cation and analysis of EMG signals on a wide variety of data set with minimum human intervention. This paper proposes a method for classi cation of ALS based on evaluation of multiple features extracted from three domains of EMG signal: time domain representation, frequency domain representation and Muscle Unit Action Potential(MUAP) waveform acquired via EMG decomposition of the signal. 43 features were evaluated using feature selection techniques like chi-squared test and recursive feature elimination. Our experimental results show that amplitude, duration and area of the MUAP waveform estimated for each motor unit, inter-spike-intervals of the motor units, variance, zero crossings, zero lag of autocorrelation, waveform length and slope sign change of the time domain representation, average spectral amplitude, total power, variance of central and mean frequency from feature domain representation of the signal provides the best accuracy at an average rate of 85%, a true positive rate(TPR) of 86% and a false positive rate(FPR) of 20% approximately.
Identifying the best metrics to find the best quality clusters of genes from gene expression data
(BRAC University, 2019-04) Choudhury, Joydhriti; Roshni, Tanzima Rahman; Chowdhury, Md. Tawhidul Islam; Rayon, Raihanoor Reza; Mottalib, Md. Abdul; Ajwad, Ajwad
Microarray data is used to create groups of similar genes based on their phenotypic attributes. Information extracted from these groups of gene can be applied to path- way analysis, disease predictions, target identification in drug design and many other important applications and functionalities in biology. However, how to determine a distance metric to measure the similarities among genes has always been a great chal- lenge. In our work, we have studied sixteen combination of distance-linkage combina- tional metrics and tried to and the groups of similar genes based on their expression level by building phylogenetic tree. Furthermore, to validate our endings we have evaluate the output of the same trails on three different datasets. Our work suggests that, Maximum distance metric with the combination of Average linkage metrics gives the optimal quality while grouping similar genes together by building a phylogenetic tree.
Detection of handwritten text using convolutional neural network
(BRAC University, 2019-04) Jasim, Rabib Bin; Mahin, Rokeya Sultana; Uddin, Jia
Machine replication of human functions, like reading, is an ancient dream. However, over the last five decades, machine reading has grown from a dream to reality. We have tried to make it more obvious through a hand writing recognition system. This research paper describes a text-line extraction based method. It offers a new solution to traditional handwriting recognition techniques using concepts of Deep learning and computer vision. An image can have hand writing, typed letters, different characters and other images. Our intention is to detect all the characters and display them. Some images can also have unnecessary lines or unclear letters. This system will clear the picture through pre-processing system and will be able to identify the letters or characters. It will help people to identify any unclear messages. It will also avoid unnecessary images and will focus on the text only. Sometimes we want to ignore unnecessary advertisement images from the newspapers. Our system will do a great work for this. It will clear all the images and unnecessary lines etc. and will only display the text what people want to read.
EEG signals analysis for motor imagery brain computer interface
(BRAC University, 2019-08) Rahman, La z Maruf; Alam, Zawad; Rahman, Md. Musta-E-Nur; Parvez, Mohammad Zavid
A brain{computer interface is a medium for communication which converts neuronal signals into commands towards controlling external system. This thesis presented the process of classifying three motor imagery tasks using EEG signals which can be further evolved into BCI system that can remotely control external devices. Different bands are ltered from EEG signals in order to extract di erent frequency distributed features. These features are used to classify di erent motor imagery tasks based on SVM and ANN. Experimental results show that SVM carried higher accuracy (i.e., 80%) compared to other machine learning algorithms where seven subjects participated in this experiment.
Emotion recognition using EEG signal and deep learning approach
(BRAC University, 2019-08) Islam, Sayedi Hassan Bin; Mehdi, Md. Quamar; Rohan, Bhuiyan Yash; Mahmood, Syed Atif Imtiaz; Parvez, Mohammad Zavid
Emotion is a mental state, which originates in the brain and is closely related to the nervous system. Emotion can be defined as a feeling expressed through, or detectable by voice intonation, facial expression body language, as response from one’s mood relationship with others and most importantly the circumstance they are in. Although, Brain Computer Interface (BCI) are being developed to find a better human-machine interaction system using brain activity and it is frequently implemented by Electroencephalogram (EEG) signals. EEG is a well established approach to measure the brain activities which can be analyzed and processed to distinguish different emotions. In this thesis, we present an approach to classify human emotions using EEG signal by Convolutional Neural Network(CNN). In our model, we use the Dataset for Emotion Analysis using Physiological signals (DEAP) dataset, a benchmark for emotion classification research, to transform the EEG signal from time domain to frequency domain and extract the features to classify the emotions. Emotion can be classified based on the two dimensions of valence and arousal. Previous researches have used fewer channels and participants. Our approach which was carried out on 32 participants, has achieved an accuracy of 94.75% for the valence and 95.75% on the arousal detection, which is quite competitive with other methods of emotion recognition.
Stock price prediction using time series data
(BRAC University, 2019-08) Mazed, Mashtura; Majumdar, Mahabub Alam
Researchers has taken a lot of years to make algorithms fast and accurate enough to make stock price predictions accurately. Investors are looking for smarter techniques to forecast stock prices for investments and this has made this topic one of the most worked out researches in data science eld. One of the trendy ways of forecasting is time series analysis. In this thesis, I have compared recent 3 most common time series forecasting algorithms that are- Autoregressive Integrated Moving Average, Facebook prophet and Long Short Term Memory, using company data (LMT and NOC) from yahoo nance. Firstly, I used K-Means clustering to choose a cluster with least number of companies and then used processed data to compare the accuracy of the algorithms.
Dhaka Stock Market analysis with ARIMA-LSTM Hybrid Model
(BRAC University, 2019-08) Arnob, Raisul Islam; Alam, Rafatul; Alam, Alvi Ebne; Majumdar, Mahbubul Alam
This paper proposes the forecasting of correlation coe cients of Dhaka Stock Ex- change market assets required for portfolio optimization using an ARIMA-LSTM hybrid model. We have developed a robust model that encompasses both linearity and non-linearity within the datasets of the Dhaka stock market with a hybrid com- bining ARIMA model and a Recurrent Neural Network called LSTM. Our hybrid model tries to utilize the unique properties of both the ARIMA model and the LSTM model. We have ltered the linear components in the datasets using the ARIMA model and passed the residuals obtained onto the LSTM model which deals with the nonlinear components and random errors. We have compared the empirical results of this model with several other traditional statistical models used in portfolio man- agement namely the Single Index model, Constant Correlation model and Historical Model. We have also predicted the correlation coe cients using the ARIMA model to see how one of the model in our hybrid performs individually. The test results show that the hybrid model excels the other models in accuracy and indicates that the ARIMA-LSTM hybrid model can be an e ective way of predicting correlation coe cients required for portfolio optimization.
Road sign detection and translation in Bangla using image processing and machine learning
(BRAC University, 2019-08) Islam, Khalid Amirul; Mubin, Nadia Farha; Zaman, Saima; Alam, Md. Ashraful
For the drivers while driving, road signs play very important role but it is really challenging for the drivers' if they are not at home with the meanings of the signs. Driving is a process which is complex, continuous, and multitasking that involves driver's perceptive ability and motor movements. Road and tra c signs and vehicle information is presented in such a manner that distracts driver's attention intensely with increased mental workload leading to safety concerns. Road sign recognition and translation system is basically a system for intelligent vehicle which guides the driver to obey the tra c rules. These are the tra c rules which are represented in a small pictorial form, erected at road sides. Our proposed work represents the process of recognizing the captured sign and then translate the detected sign in Bangla. Translation of an image is done by matching multiple images. We are presenting a system for detecting and recognizing the signs displayed around us and voice synthesizing their contents with an algorithm for detection and recognition of road sign using image processing and machine learning method. We have used a template matching algorithm to match the road signs with the database after recognizing road signs it will be translated in Bangla or English according to user preference and manufactured as voice output note. The detection, recognition and speech synthesis modules each will perform their respective tasks e ectively and e ciently, and the future advancement of the currently proposed system is promising.
Bangla sign language recognition using leap motion sensor
(BRAC University, 2019-08) Tan, Tamkin Mahmud; Mondol, Anna Mary; Nawal, Noshin; Ahmed, Sabbir; Uddin, Jia
Sign language is used by hearing and speech impaired people to transmit their messages to other people but it is difficult for a regular people to understand this gesture based language. Instantaneous responses on sign language can significantly enhance the understanding of sign language. In this paper, we propose a system that detects Bangla Sign Language using a digital motion sensor called Leap Motion Controller. It is a sensor or device which can detect 3D motion of hands, fingers and finger like objects without any contact. A Sign Language Recognition system has to be designed to recognize a hand gesture. In sign language system, gestures are defined as some specific patterns or movement of the hands to give an expression. There has to be a library which includes all the datasets to match with the user given gestures. We have to compare the sequences of data we get from Leap Motion and our datasets to get an optimal result which is basically the output. It will then show the output as text in the display. For our system, we choose to use $P Point-Cloud Recognizer algorithm to match the input data with our datasets. This recognition algorithm was designed for rapid prototyping of gesture-based UI and can deliver an average over 99% accuracy in user-dependent testing. Our proposed model is designed in a way so that the hearing and speech impaired people can communicate easily and efficiently with common people.
Smart surveillance system for identifying bikers without helmets using deep learning
(BRAC University, 2019-08) Hossain, MD. Iqbal; Muhib, Raghib Barkat; Chakrabarty, Amitabha
Modern world is progressing quickly along with technology and one of the major sectors is transportation technology. Day by day the number of people are increasing and the number of vehicles are increasing too. As a result we have easy access to vehicles these days but at the same time it has increased the number of road accidents. Among the various types of road accidents, motorcycle accident is one of the main type which causes severe injuries and in some cases death. The only protection motorcyclists use is their helmet. Most countries has a law on wearing helmet otherwise it will be punished, but many people often break this law and as a result it increases the percentage of severe injury and death. In a populated country it's hard to keep track of the bikers who don't use helmets because of the huge number of bikes moving at a time. In this circumstances, we have developed a solution which can identify the bikers who don't use helmets. Our system uses image processing and deep Convolutional Neural Networks(CNNs) which is used to identify who breaks the law of wearing helmet, with helmet vs without helmet identification and finally motorcycle license plate recognition. We have used multiple models of object identification and evaluated in terms of speed vs accuracy. The results we have got indicates that due to the increase of law enforcement and awareness of the traffic police, the use of helmet has decreased a bit than before but there are still a lot of bikers who don't use helmets on regular basis. As we are going to detect people without helmet and then will keep the registration number from the license plate, we can easily identify the law breakers and the government can punish them accordingly. We have used the Tensor ow library for our system. The models we have used in our system are SSD Mobilenet v2 and Faster RCNN inception v2. For SSD Mobilenet V2 the accuracy for the helmet was 90 percent, human was 55 percent, bike was 80 percent and number plate was 95 percent. This model is quite light and we can use this trained model even in mobile devices. For Faster RCNN inception v2 it took more computations but the accuracy we got was slightly better as it is more heavyweight than SSD Mobilenet. The accuracy we got for this was 92 percent for the helmet, 58 percent for the human, 81 percent for bikes and 96 percent for number plates.