Search

Facial expressions depict emotions and produce information on the personalities and thoughts of people. The machine performs different tasks constantly in order to increase its use in public. Machines that are able to understand emotions can be used to execute a wide range of tasks. Machine perception demands machines to grasp information about their environment. The understanding of human facial expressions is a key component to understand emotions and also to find broad applications in the field of human-computer interaction. Tensorflow framework, Keras library and the Xception Architecture of CNN are used to train the model on the Fer2013 dataset.

Parametric classification fits the parametric model to the training data and interpolates to classify the test data, whereas nonparametric methods like regression tree and classification trees use different techniques to determine classification. The classification process can be of two types: supervised and unsupervised. In supervised classification, training data are used to design the classifier. Bayes’s rule, nearest neighboring rule, and perceptron rules are few widely used supervised classification rules. For unlabeled data, the process of classification is called clustering or unsupervised classification. This paper proposes a wrapper-based approach for pattern classification to minimize the error factor. Techniques, such as Bayes’s classification, K-NN classifier, and NN classifier, are used to classify the patterns using linearly separable, linearly nonseparable, and Gaussian sample dataset. These methods classify the data in two stages: training stage and prediction stage. In this paper, we will be using parametric and nonparametric decision-making algorithm as we know the statistical and geometric properties of the patterns under study.

Our project aims at building a biometrics based attendance registry system for educational institutions such as colleges and schools. We have tried to bring together concepts of hardware and software engineering to give an end consumer product which can replace the current manner of attendance marking. This project also utilizes the fundamentals of Internet of Things (IoT) for data transfer, storage and display. The system can be interpreted based on it's hardware and software aspects. The hardware section consists of a pod, which is a portable device that can be circulated among students in the classroom. It features an LCD screen that displays various functional options and a, which can scan, store and identify student fingerprints. The identified fingerprint ID is then stored in memory and is made ready for transfer. The data is then sent to the database on a web server where the Database Management System (DBMS) updates the attendance tables accordingly. This database can be downloaded or viewed by the respective faculty.

This paper deals with the new segmentation techniques for retinal blood vessels on fundus images. This technique aims at extracting thin vessels to reduce the intensity difference between thick and thin vessels. This paper proposes the modified UNet model by incorporating ResNet blocks into it which includes structured prediction. In this work we generate the visualization of blood vessels from retinal fundus image for two loss functions namely cross entropy loss and Dice loss where the network classifies several pixels simultaneously. The results shows higher accuracy by considering a much more expressive UNet algorithm and outperforms the past algorithms for Retinal Vessel Segmentation. The benefits of this approach will be demonstrated empirically.