Advances in Science and Technology Vol. 124

Abstract: Facial recognition based music system plays an important role in the treatment of human psychology. Face recognition system is an extensively used technique in most of the applications such as security system, video processing, in surveillance system and so on. People are often confused while choosing the kind of music they would want to listen. Relatively, this paper focuses on making an efficient music recommendation system which will recommend a suitable music to make the person feel sooth using Facial Recognition Techniques. This system uses FER-2013 dataset for training of the CNN, which is made using mini-xception architecture. Augmentation techniques are used for increasing the number of images in the dataset for training, which helps to increase the accuracy of the prediction. The face is captured using webcam and facial extraction is done using Haarcascade classifier and then sent to the CNN layers. The mini xception algorithm used in these CNN layers makes the system lighter and efficient as compared to existing systems. The accuracy of the proposed model is calculated and found to have reached the barrier threshold of 95% and average accuracy was found to be 90%. The song is recommended to the user using the proposed mapping algorithm.

Abstract: Machine literacy refers to the creation of digital mechanisms which could make significant contributions without formal training by interpreting as well as extrapolating from attack patterns using mathematical techniques from various models. It's research wherein the system can improve itself autonomously as a consequence of the interaction and data. It's an example of artificial intelligence in action. In this study, we'll look at recognizing bike riders wearing or not wearing helmets in a video. To get the movies back into circulation, we'll have used the OpenCV Software. The YOLO and CNN models, that are utilized for enabling real-time license plate retrieval for a non-helmeted rider, the above two models are indeed the prescribed strategy for our device. The technology in our approach looks to see if the user who is riding the bike is wearing a helmet. If the rider is not wearing the required protective gear, the license plate is nonetheless extracted so that it could be supplied to activity recognition technology for evaluating and giving the fines to the prosecutor. For this design, we have a truly vast compass. It can be used to detect bikers who are not wearing a helmet. If the individual in the video is not wearing a helmet, the system is designed to detect the license plate of the bike and send an appropriate alert message to the appropriate person. This system is found to be more robust and effective than other algorithms.

Abstract: The primary way to classify retinal illnesses is to conduct several medical examinations, the most important of which is a visual examination. Human error is common as a result of a poor-higher cognitive process, which is one of the major challenges in visual disease diagnosis. Automated image processing technologies are more useful for early disease diagnosis and evaluation than the digitized diagnostic imaging conventional operations are confusing and time-consuming. The aim of this paper is to create a system that detects retinal abnormalities based on images using Deep learning technique. The images are first pre-processed. The photographs are enhanced after they have been pre-processed. The images that have been pre-processed are fed into the Penta-Convolutional Neural Network (Penta-CNN). Penta-CNN is a five-layered architecture that includes two convolutions, max pooling, and three fully connected layers. The performance of Penta-CNN is evaluated using STARE(Structured Analysis of the Retina) database [14]. The model is also trained with several hyperparameters which are tweaked and assessed.

Abstract: A real-world weather prediction system that detects and describes weather condition in image data is becoming prominent subject in machine vision . These systems are designed to address the challenge of weather classification using machine vision. Advances in the fields of Artificial Intelligence and Machine Learning enables applications to take on the image recognition capabilities to identify the input image . Deep learning is a vast field and narrow focusing a bit and takes up the challenge of solving an Image Classification process. Proposed deep learning algorithms by tensorflow or keras by classifying the image of weather reports by convolution neural network. CNN is an artificial neural network which inspires animal neural cortex and built upon it . The images are passed through the neural network which consists of multiple layers and filters and then identified and classified according to the weather type. The algorithm is inspired by the brain and so named as Convolution neural network.

Abstract: COVID-19 commonly known as Coronavirus is caused by the virus SARS-CoV-2. This pandemic can be prevented by wearing a face mask after taking the vaccines that have arrived in the market, A vast trial led by researchers at Yale University and Stanford Medicine came up with the conclusion that if a person’s mouth and nose is covered by a face mask they would have a lesser chance of getting infected by COVID-19. To contribute towards global health, this project aims to develop an alert system for face mask detection in public places. Our system tries to achieve this using Keras, TensorFlow, MobileNet, OpenCV, PyTorch, and R-CNN. The proposed work also does Facial landmark detection which can be further extended in the future with face mask detection for applied in Biometric applications. This paper is aimed towards creating a real-time and highly precise technique which efficiently identifies faces that are masked, unmasked, and incorrectly masked and alerts in the case of anomalies to ensure that masks are worn properly.

Abstract: The objective of a project is to make an integrated technological solution to assist and help companies in their hiring process. We have come up with the concept for prioritization of candidates using automated behavioural interview. This is a web-based portal that will be supporting the recruitment operations by prioritizing candidates from a large pool of candidates based on skills and experiences relevant to the job. Facilitating an informed interview process for a candidate with systematic transparent procedure. Our project aims to apply our technical skills to facilitate interviewers and address the challenges faced by the organization in its hiring process. This has been addressed using recording the answers of random behavioural questions with limited time to prepare, then video will be analysed to find the behavioural aspect of the candidate. Behavioural analysis of the interview is done by applying facial emotion recognition based on VGG Net Architecture.

Abstract: The present pandemic situation has made people not go out anywhere because it’s getting difficult to learn the concepts briefly for both lecturers and students. Concepts are learned through videos, but reading is also an important aspect of learning. This paper talks about providing books and notes in online mode to read and recommend books based on facial expressions captured from the user. This paper aims to extract faces from an image, extract the expression (eyes and lips) from it and also classify them into six types of emotions, which are Happy, Fear, Anger, Surprise, Neutral and Sad. The algorithm used for facial expression recognition is the Convolutional Neural Network algorithm, also known as CNN.

Abstract: In the recent past there has been an increase in the occurrence of violent incidents involving dangerous objects such as arms and knives. Being able to quickly identify and defuse such situations are of utmost importance in order to preserve peace and to avoid human casualties. One of the most important and commonly used methods to increase security is the usage of surveillance cameras almost everywhere. The benefit of object detection techniques can be used in this field in order to help improve security. Using object detection techniques in order to detect objects of interest in surveillance footage is one method to identify dangerous situations and take necessary steps in order to minimise any damages.This paper uses convolutional neural network (CNN) based YOLO algorithm in its implementation to detect weapons such as knives and pistols

Abstract: Plants are a major and important food source for the world's population. Smart and sustainable agriculture should be capable of providing precise and timely diagnosis of plant diseases. This helps in preventing financial and other resource loss to the farmers. Since plant diseases show visible symptoms which a plant pathologist will be able to diagnose through optical observations. But this process is slow and requires continuous monitoring as well as the availability and successful diagnostic capability of the pathologist. To overcome this, in smart agriculture, computer-aided plant disease diagnostic/detection model is used to help increased crop yield production. Common diseases are found in tomatoes, potatoes and pepper plants, some of them are bacterial spot, early blight etc. If a farmer can detect these diseases early, and can apply an appropriate treatment then it will improve crop yield and prevent the economic loss. In this work, we train the dataset on three different deep convolution neural network architecture and found the best suitable model to detect tomato leaf diseases. In order to avoid overfitting of the mode, batch normalization layer and a drop out layer has been included. The proposed Deep CNN is trained with various dropout values and a suitable dropout value is identified to regularize the model. The experimental methodology tested on plant village dataset showed improved accuracy of 96%, even without performing pre-processing steps like noise removal. By introducing batch normalization and dropout layer training accuracy improved to 99% whereas validation and testing accuracy is found to be 98%.

Abstract: In the growing field of autonomous driving, minimizing human errors lead to fatal accidents. Many companies and research groups have been working for years to achieve a fully autonomous vehicle. Self-driving cars are inevitably the future. The system to be implemented is to train and directly translate images from three cameras to steering commands. The proposed method is expected to work on local roads with or without lane markings. With just the images generated and the human steering angle as the training signal, the proposed method is expected to automatically recognize important road characteristics. A simulator (Udacity) is used to generate data for the proposed model.