Potholes are a common problem in roads and highways around the world, which can cause severe damage to vehicles and create safety hazards for drivers.In recent years, deep learning algorithms have been increasingly used for automated pothole detection.This research offers a deep learningbased algorithm that can detect potholes early using photos and videos, reducing the likelihood of an accident.This model is basically based Faster Region-based Convolutional Neural Network(F-RCNN) and You Only Look Once Version 3(YOLO V3).It also discuss the challenges in detecting potholes, such as variable lighting conditions and noise in the data, and how these challenges have been addressed in previous research.Finally, we provide a comparative analysis of the performance of different deep learning algorithms for pothole detection based on accuracy.There are various pothole identification models that combine the accelerometer with machine learning techniques, but there are fewer pothole detection models that use simply machine learning techniques to detect potholes.The findings of this study suggest that deep learning algorithms can provide accurate and efficient pothole detection solutions that can help road authorities to maintain and repair roads, reduce vehicle damage, and enhance road safety.
The widespread dissemination of misinformation, commonly known as fake news, has been facilitated by the rapid expansion of social networks as platforms for news distribution.This stands in contrast to traditional mass media channels such as newspapers, magazines, radio, and television.The challenges arise from human limitations in discerning between true and false information, which poses a significant threat to logical coherence, democratic processes, journalistic integrity, and the credibility of government institutions.The lack of reliable and trustworthy information on social media further compounds the issues associated with this phenomenon.To address this pressing problem, we have proposed an integrated system that incorporates various pre-processing techniques and classification models.This system aims to detect and combat fake news by evaluating their efficacy on a specific dataset of labeled news statements.By utilizing metrics such as precision, F1 scores, and recall, we can determine the most effective model.The primary objective of this system is to develop an efficient and accurate model capable of predicting and identifying instances of fake news within social media networks.
Software Engineering is a Comprehensive domain since software system have become larger and complex than ever.Such software characteristics make it very complex to develop fault free software.Therefore, identifying defects automatically and fixing them is challenging task.Improper modelling, lack of requirement specifications, wrong coding, poor configuration management may cause defects which leads to failure of software system.Such defects must be detected and fixed at early stage of software development to minimize cost.Machine learning algorithms widely used in the software defect prediction also achieves good results in predicting software defects using deep learning techniques.In this paper, we are providing comparative study of various algorithms like convolutional neural networks, multi-layer Perceptrons in identifying defects in software.Moreover, the experiments conducted on NASA datasets.
Wouldn't it be cool for your device to translate English language into the language you speak in real life and vice versa?Understanding multiple languages can be a challenging task for humans.In a country like India, where there are over 22 recognized languages, it can be nearly impossible for an individual to learn and comprehend all of them.Consequently, communicating effectively in such a diverse linguistic environment can pose significant obstacles, forcing the need of human interpreters.An interpreter is a platform which translates what is being said in one language into another language without changing its original meaning.Google Translate is one such existing platform but it has drawbacks like not ideal for confidential documents, cannot provide a perfectly accurate translation and requires internet connection to translate.Our goal is to address the limitations mentioned earlier, and to accomplish this objective, we have come up with a project idea that involves real-time translation of spoken languages.The project aims to achieve this by using audio as an input.The proposed system will take real-time English audio as an input, translate it into Telugu and then will produce audio output in Telugu and vice versa.As English is the third most spoken native language in the world after Chinese and Spanish and also the most widely used, we chose to convert the text from English to Telugu and vice versa.Our intention is to utilize a Sequence-to-Sequence model as it is proficient in performing language translations.This model works by transforming a sequence of words from one language into a sequence of different words in another language.It uses neural network for sequence learning.
Software companies will constantly work on numerous projects at a time. Project managers are responsible for making each project successful. As to monitor and control many projects organizations use project management tools to perform tasks in a well-organized manner. Current day’s lot of data is generated in organizations related to software development as the organizations use project management tools to track the status of the project. This project data generated by project management tools like Jira, Asana, and Trello can be used to solve resource allocation problems. Though there are numerous solutions for addressing the project scheduling problem, none of them has intelligent decision-making. Research is still scarce in the area of software project management and evidence is required to assess the theoretical methods and ideas. This paper introduces a resource allocation method that uses machine learning for effective project scheduling.
In a day and age where billions of users regularly use social media and express their opinions online, there is potentially a lot of data that can be harvested and utilized; therefore, it is crucial to develop a quick way to garner data. This study aimed to develop a program using Python to do the same, trying to understand the sentiments of the authors of the text and tweets as well as other additional information about the top tweets and retweets. The main objective of the Twitter Sentiment Analysis is a query-based analysis of tweets. In simple words, Twitter Sentiment Analysis focuses on analyzing the tweets of a specific/particular topic that the user wants to analyze. An extensive collection of such sentiments could leverage to provide a fair reflection of public sentiment towards a specific topic. There are thousands of tweets that can be quickly processed for the sentimental impact, compared to the amount of time it would take a large team of people to complete the same task manually. There are tons of text documents that can also be processed for sentiments in seconds, much faster than just a team of people manually skimming through the text.
The increasing number of Internet users and digital immigrants has led to increase in enormous types of queries asked by the users. For finding the best results of the queries, abundant search engines and meta-search engines are being used with different and efficient result providing features. The existing meta-search engines uses various search engines for fetching the results but, do not emphasize on the semantic analysis of the query for identifying the best search engine suitable for the query of user. In order to overcome this limitation, a meta-search engine is proposed. The proposed meta search engine can improve quality of results through the use of semantic analysis and query processing. The end results comparison with existing meta search engine proves the proposed approach better than the existing meta search engines.
Computer Vision is applied in agriculture field for food grading, disease identification of the plants and agro-farms security.Huge crop damage is caused by the wild animal attacks on the agriculture farms.Here are some traditional techniques followed by the local farmers, but which are not effective.This problem can be solved using computer vision techniques.In this paper, we proposed an algorithm to detect animals in a given image.W-CoHOG is a Histogram oriented gradients based feature vector with better accuracy.It is an extension of Co-occurrence Histograms of Oriented Gradients (CoHOG).In this paper LIBLINEAR classifier is used in order to get better accuracy for high dimensional data.The experiments were conducted on two benchmark datasets called Wild-Anim and CamaraTrap dataset.Experimental results prove that W-CoHOG performs better than existing state of the art methods
In present scenario, where most of the parents are working persons, they cannot be around their child all the time. Hence most of the time, the child is either kept with a daycare or childcare facility or a maid is kept at home for taking care of the child. Recent newspaper articles have reported lots of incidents on child abuse, mostly in daycare facilities or by maids at home. Hence the requirement of remote surveillance system is the need of the hour. A major disadvantage of a general surveillance system is that they can only monitor a limited area in room and hence creates blind spots. There is no automatic movement of camera and it needs to be adjusted manually to cover a different angle. In the proposed system, we are using Raspberry Pi microcomputer and a camera, which is made dynamic using Passive Infrared (PIR) sensors and Servo motor. The camera will rotate according to the movement of the child around the room it covers. The camera will be installed in the centre of the room so that it can cover all directions .The video captured can be streamed live online and parents can access the feed by logging in to a website. A GSM module has been used, so that the parents can get SMS alert whenever any sensor is activated. The proposed setup is a low cost surveillance system and can be implemented at home or childcare facilities.
In recent years storing the visited solutions during evolutionary process is becoming one of leading topic in the area of research which guarantees that the algorithm will never get trapped to local optimum. This paper focuses on solving the min cut balanced partitioning problem for digital circuit layout using a solution archive which stores efficiently all visited solutions during the evolutionary process and at same time intelligently transforming the revisited solutions into yet unvisited feasible solutions giving balanced weighted partitions.
In todays' world, the latest cognitive technologies available in markets are not comprehensible to all the individuals.However some technologies are available that helps all the individuals to be more interactive with others in the society thereby reducing the technological divide.Some of the products are there like audio navigation system for blind people or video navigation system for deaf people or voice command wheel chair for physical disabled people etc. are enabling them to live better without any help of any other fellow human being.Most of the cases consider either lesser interactive device with less accessibility or highly cost effective tools have been used.There is lacks of cost effective intelligent interactive systems for an individual.This study proposes to develop a holistic approach using Human Computer Interaction to design an interactive system that minimize the barrier between the human's cognitive model of what they want to accomplish and the computer's understanding of the user's task by integrating techniques or applications to enable an individual to use the advanced technologies in more interactive manner.It is also planned to develop a prototype with minimal complexities and by retaining and enhancing the required features, thereby providing a supportive system that can meet the need of its deliberate users.
In the field of medical diagnostic, magnetic resonance (MR) image is an important and popular medical image, which technically required enormous data to be stored and transmitted. In order to make accurate diagnostic for the patients those data from the MR images need to be high accuracy and completeness. Various algorithms have been proposed to improve the performance of the compression scheme. One of our contributions in this paper is to demonstrate the choice of decomposition level is playing a very important role in achieving superior wavelet compression performances. We extended the commonly used algorithms to image compression and compare its performance. For the best image compression performance, lifting based Cohen-Daubechies-Feauveau wavelets with the low-pass filters of the length 9 and 7 (CDF 9/7) wavelet transform is used, which coupled with Set Partition in Hierarchical Trees (SPIHT) coding algorithm and entropy coding techniques. The final simulations showed that all those used technologies have made a large reduction of image size occur. It needs to be highlighted that the compression ratio has been significantly improved by 99%, together with highest PSNR values and MSSIM by overall recognition rate 96.17%.
Risk analysis is the way of preventing the damage by the prevention mechanism. The proper risk analysis techniques employed will reduce the vulnerabilities, threats, impacts to a great extent. Here the risk management of information security is dealt with the anomaly detection mechanism. The Anomaly detection is carried out by the data clustering to find the outlier as the anomaly. The clustering process is enhanced by the blended mechanism of genetic algorithms and the Ant Colony Optimization. The quality of the clusters obtained are improved and demonstrated with the results. The proposed Anomaly detection process is compared with two other methods and the results obtained are appreciable.
Feature generation algorithms for searching globally useful features using traditional Markov network structures is now a days in wide practice. The composition of a Markov network can be represented one of two ways. The first approach is to treat this task as a global search problem. However, these algorithms are slow as they require running the expensive operation of weight learning many times. The second approach involves learning a set of local models and then combining them into a global model. However, it can be computationally expensive to learn the local models for datasets that contain a large number of variables and/or examples. In this paper, we follow a third approach that views Markov network structure learning as a feature generation problem. For this we used an algorithm DTSL (Decision Tree Structured Learner) which combines a data-driven, adhoc-to-generic search strategy with randomization for quickly generating a large set of candidate features that all have support in the data. In addition to that it uses weight learning, using forest of uncertain decision trees to select a subset of generated features for making feature generation process more accurate and effective.
Augmented Reality & Mobile Augmented Reality has very widely used in nowadays Smartphone's and there is not any Software Architecture available. So when one application will develop for Mobile Augmented Reality will targeted only on one mobile handsets or device. When the device will change for some reason the application will require to Re-Architect because of its operating system and handle the network and GPS. Also the application will required to Install in device and require some kind of storage and some extra libraries to developed Mobile Augmented Reality System. In this paper propose algorithm for web based service oriented architecture for mobile augmented reality system. Using this prototype algorithm user can design any mobile augmented reality system application; also extend the services using current web services.A Framework is proposed in order to achieve any Application design using this Software Architecture will run on any mobile device platforms without the need for Installation or no Installation. Architecture will give the "Web Based Interface" so the User's have some kind of Network support and GPS Support will easily get the Application on any mobile Device.Also the framework will base on SOA so web services will design according to handle the Mobile data e.g. Images, Audio, and Video etc. to and from the web server. Software Architecture will give the Application to Portability, extensibility, No Need for Installation, Reusability without any code change. Same code will run on multiple mobile devices without the OS limitation and any third party will design his own application using the same code or extend it as per his choice.
This paper presents a study of various genetic algorithms developed for solving the Job shop scheduling problem. Various approaches such as crossover operators, mutations and constrained problem statement have been applied to obtain optimal solutions. Some of the key areas studied are: , penalty function, VND[Variable Neighborhood Descent Algorithm], random keys, JOX[ job based order crossover], GOX[generalized order crossover] and OBGT[order-based Giffler and Thompson].
Web services play an important role in the enhancement of the business in the modern world. The impact of non-functional requirements on web engineering principles and practices may dictate the need of creation of novel techniques and tools and the enhancement of existing tools and techniques. In this paper we have device the method to improve the performance by the web personalization process. The web personalization is done by the usage mining. Association of the transactions paves the way for the personalization and in turn the performance improvement. Here the Apriori algorithm is chosen for the deployment, the rules generated are optimized by the genetic algorithms and discussion for the choice is also made based various metrics.
