Advances in Science and Technology Vol. 124

Abstract: This paper gives a new technique for video editing by introducing modules that are trained to create realistic and dramatic sky backgrounds in videos. This project is different from what is being used in the video editing ecosystem as it does not require any static photos or inertial measurements. The module can be simply used on any device without having any prerequisites in the device. This is a game- changer when it comes to capturing cinematic sky videos. This project is further branched into three different modules to segregate the different tasks including sky routing, flow reckoner, and image emulsifier. These methods will run in real- time and are user friendly. This project can generate high fidelity videos with different lighting and dramatics in outdoor environments. Adding further we can also easily synthesize different weather conditions. This editing technique is much simpler and easier giving a more aesthetic image for cinematic shots.

Abstract: Nowadays face recognition system is widely used in every field of computer vision applications such as Face lock-in smartphones, surveillance, smart attendance system, and driverless car technology. Because of this, the demand for face recognition systems is increasing day by day in the research field. The aim of this project is to develop a system that will recommend music based on facial expressions. The face-recognition system consists of object detection and identifying facial features from input images, and the face recognition system can be made more accurate with the use of convolutional neural networks. Layers of convolution neural network are used for the expression detection and are optimized with Adam to reduce overall loss and improve accuracy. YouTube song playlist recommendation is an application of a face recognition system based on a neural network. We use streamlit-webrtc to design the web frame for the song recommendation system. For face detection, we used the Kaggle-FER2013 dataset, and images in the dataset are classified into seven natural emotions of a person. The system captures the emotional state of a person in real-time and generates a playlist of youtube songs based on that emotion.

Abstract: Driver weariness is the leading cause of road accidents, according to numerous studies. Computer vision algorithms have shown promise in detecting indicators of exhaustion from facial motions like yawning. Precision and consistent yawning recognition is difficult in the real-world driving environment due to the various facial gestures and expression of driver. Yawning causes a mouth deformation in a variety of facial activities and expressions. This paper provides a novel way to based on minor facial motion identification to address the aforementioned concerns. We offer a for nuanced face activity recognition. Bidirectional and 3D convolutional networks are used in this network. A keyframe selection technique is used to discover the most from delicate face gestures. This method employs photo histograms to swiftly eliminate redundant frames and the median absolute deviation to locate outliers. A variety of tests are also run on the method.

Abstract: Alzheimer’s disease is a delusional brain syndrome that causes hindrance in the functional ability of a person. This is progressively marked by shrinking of the brain and continuous loss of brain cells. Consequently, it leads to death and thus it becomes important to come up with a system that can catch this disease early on. MRI (Magnetic Resonance Imaging) has evolved into a valuable medical diagnostic tool for the diagnosis of brain and other medical imaging over time. In the past a lot of data has been collected by different researchers and a variety of machine learning algorithms have been used to diagnose this disorder and label it into different classes. Through this project we are presenting a CNN based model trained on MRI images to diagnose this disease effectively. The use of CNN is a no-brainer as apart from being an excellent classifier, it is a very good feature extractor which reduces the overall cost of feature engineering. The proposed model takes an MRI image as input and classifies it into very mild, mild, moderate or no disease categories. The trained model has a 95 percent accuracy rate.

Abstract: Diabetic Retinopathy (DR) is a condition in which damage to the eyes occurs as a result of diabetes mellitus. It is the most frequent diabetes-related eye condition. It can also cause full blindness and vision loss. With effective eye treatment, the majority of new occurrences of diabetic retinopathy can be reduced. Early detection helps to avoid total vision loss. However, detecting it early can be difficult because it may not present symptoms in the early stages. The wide selection of fundus imaging makes classification challenging, mainly in Proliferative_DR, which includes the formation of new vessels in retina and bleeding. Pre-trained deep learning model is used on the publicly accessible retinal fundus image dataset on kaggle in this paper (APTOS 2019 Blindness Detection). Pre-processing and augmentation procedures are used to increase the accuracy of the models that have been pre-trained. The training accuracy of 8-Layer Convolutional Neural Network (CNN) and MobileNetV2 obtained is 83.07% and 85.21%. Testing accuracy achieved 71.93% using CNN & MobileNetV2 is 83.42%. The most often employed measures, such as the F1 Score, precision, and recall is used to ignore class level of label disagreement, which aids in diagnosing all phases of diabetic retinopathy. The results using a confusion matrix is analyzed, which is useful for categorising different stages of diabetic retinopathy according to severity. It also takes into account the degree of mismatch between the actual and anticipated labels.

Abstract: Smart farming is an innovative technology that aids in the improvement of the country's agricultural produce quality and quantity. Wheat is the most important crop in most parts of India. Wheat leaf diseases have a significant impact on production rates and farmer earnings. It poses a significant danger to food security because it affects crop productivity and degrades crop quality. Accurate and precise disease detection has posed a significant challenge, but recent advances in computer vision enabled by deep learning have paved the road for camera-assisted wheat leaf disease diagnosis. Using a CNN trained with a publicly available wheat leaf disease model, several machine learning algorithms and neuron- and layer-wise visualization methods are applied.

Abstract: There has been a vast development in the field of the internet, it has led people to express their opinions. It is very important to understand the customer for the successful businesses. Customers express their thoughts in the form of reviews, which can be positive or negative. For successful businesses, it is essential for them to understand a customer and their behavior because it will help them to grow their business more successfully. In this paper, we have proposed sentiment analysis of restaurant review datasets using multinomial naive bayes and logistic regression. This program will help owners quickly determine the customer's sentiments.

Abstract: In the field of Natural Language Processing, Question Answering is a cardinal task that has garnered a lot of attention. With the development of multiple language models, question answering systems have been developed and deployed to facilitate enhanced information retrieval. These systems, however, have been implemented to a large extent only in English. Our objective was to create such a question answering system for the Tamil Language. We decided to use XLM-RoBERTa as our language model, which has been trained on a variety of datasets. We have also employed a hand-annotated dataset for the purpose of validation. We trained the model on two types of datasets, the first one being only in Tamil, whereas the other one being a mixture of Indian languages along with Tamil. The results were satisfactory in both cases. Given the huge amount of computational power the model required for training, we utilized the Colab Pro Plus cloud GPU from Google to satisfy our demands. We will also be publishing our dataset on huggingface so that fellow researchers can use it for further analysis.

Abstract: Speech Emotion Recognition, as described in this study, uses Neural Networks to classify the emotions expressed in each speech (SER). It’s centered upon concept where voice tone and pitch frequently reflect underlying emotion. Speech Emotion Recognition aids in the classification of elicited emotions. The MLP-Classifier is a tool for classifying emotions in a circumstance. As wave signal, allowing for flexible learning rate selection. RAVDESS (Ryerson Audio-Visual Dataset Emotional Speech and Song Database data) will be used. To extract the characteristics from particular audio input, Contrast, MFCC, Mel Spectrograph Frequency, & Chroma are some of factors that may be employed. To facilitate extraction of features from audio script, dataset will be labelled using decimal encoding. Utilizing input audio sample, precision was found to be 80.28%. Additional testing confirmed this result.

Abstract: Social media is the highly gained popular medium of interaction.And because of using the social media in our day to day life more oftenly,the increase in cyberbullying has also increased,especially to the young people who are more into the social media platform.So by cyberbullying detection technique we can make a world or environment more safe.cyberbullying detection technique will categorize tweets or comments as bullying and non bullying words. The previous existing system uses algorithms such as Naïve Bayes which is slow and less accurate in identifying cyber bullying and also high false positives is observed.Instead here CNN and LSTM is proposed to identify cyberbullying comments and to provide good amount of accuracy